Datasets:

SouthernCrossAI
/

Arxiv

Dataset card Data Studio Files Files and versions Community

Split (1)

train · 132k rows

Filename stringlengths 22 64	Paragraph stringlengths 8 5.57k
Processed_Three_flavor_dynamical_QCD_project_by_CP-PACS_JLQC.txt	lations of mq. In our previous studies of quenched and two ﬂavor full QCD, however, we have shown that they converge to universal values in the con- tinuum limit [3,4].
Processed_Three_flavor_dynamical_QCD_project_by_CP-PACS_JLQC.txt	shows a smaller scaling violation than VWI quark masses [4]. To convert to the quark mass in the MS scheme, we match the renormalized mass with the lattice data at µ = a−1 using a meanﬁeld- improved one-loop Z factor [21], and let it run to µ = 2 GeV using the 4-loop beta function.
Processed_Three_flavor_dynamical_QCD_project_by_CP-PACS_JLQC.txt	Our results are summarized in Fig. 7. Previous results from quenched and two ﬂavor full QCD, extrapolated to the continuum limit, are also shown for comparison [3,4]. We see that quark masses decrease when we increase the number of dynamical quark ﬂavors. In three ﬂavor QCD, we ud = 2.89(6) MeV and mMS obtain mMS s = 75.6(3.4) MeV at a−1 ≃ 2 GeV, where the central values are from the K-input and the errors include system- atic errors estimated by the diﬀerence between the K- and φ-inputs. With the eﬀects of the dy- namical s quark, both mud and ms are decreased by about 15% from the previous two ﬂavor val- ues. The ratio ms/mud = 26.2(1.0) is consistent with the one-loop estimate of chiral perturbation theory 24.4(1.5) [22].
Processed_Three_flavor_dynamical_QCD_project_by_CP-PACS_JLQC.txt	I thank the members of the CP-PACS/JLQCD Collaborations for discussions. This work is in part supported by the Large Scale Simulation Program No. 98 (FY2003) of High Energy Accel- erator Research Organization (KEK), by Large Scale Numerical Simulation Project of S.I.P.C., Univ. of Tsukuba, by Earth Simulator Research Projects, ES Center, and by the Grants-in-Aid of the Ministry of Education (Nos. 1315204 and 13640260).
Processed_Three_flavor_dynamical_QCD_project_by_CP-PACS_JLQC.txt	the CP-PACS com- puter, Nucl. Phys. B (Proc. Suppl.) 60A (1998) 246. See also http://www.rccp.tsukuba.ac.jp/.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	ABSTRACT We present 21 new radio-continuum detections at catalogued planetary nebula (PN) positions in the Large Magellanic Cloud (LMC) using all presently available data from the Australia Telescope Online Archive at 3, 6, 13 and 20 cm. Additionally, 11 previously detected LMC radio PNe are re-examined with 7 detections conﬁrmed and reported here. An additional three PNe from our previous surveys are also studied. The last of the 11 previous detections is now classiﬁed as a compact H ii region which makes for a total sample of 31 radio PNe in the LMC. The radio-surface brightness to diameter (Σ–D) relation is parametrised as Σ ∝ D−β. With the available 6 cm Σ–D data we construct Σ–D samples from 28 LMC PNe and 9 Small Magellanic Cloud (SMC) radio detected PNe. The results of our sampled PNe in the Magellanic Clouds (MCs) are comparable to previous measurements of the Galactic PNe. We obtain β = 2.9 ± 0.4 for the MC PNe compared to β = 3.1 ± 0.4 for the Galaxy. For a better insight into sample completeness and evolutionary features we reconstruct the Σ–D data probability density function (PDF). The PDF analysis implies that PNe are not likely to follow linear evolutionary paths. To estimate the signiﬁcance of sensitivity selection eﬀects we perform a Monte Carlo sensitivity simulation on the Σ–D data. The results suggest that selection eﬀects are signiﬁcant for values larger than β ∼ 2.6 and that a measured slope of β = 2.9 should correspond to a sensitivity-free value of ∼ 3.4. Key words: planetary nebulae: general – Magellanic Clouds – radio-continuum: galaxies – radio-continuum: ISM – planetary nebulae: individual.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	seven previously reported RC detections in the LMC, and three PNe which were detected in LMC surveys and reported in Filipovi´c et al. (2009).
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	The Australia Telescope Compact Array (ATCA) archive databases contain unprocessed Australia Telescope National Facility (ATNF) radio astronomy data obtained with the ATCA and other Australian facilities managed by the Com- monwealth Scientiﬁc and Industrial Research Organisation of Australia (CSIRO). This data is accessible through the Australia Telescope Online Archive (ATOA) search page1.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	We ﬁrst searched the ATOA to select any projects with data at 3, 6, 13 or 20 cm that were obtained using a 6 km conﬁgu- ration of the ATCA antenna array in order to select projects with the best possible resolution. Typically, these projects contained raw data for multiple images with diﬀerent cen- tral coordinates. A project was selected for processing if an LMC PN was located within a 10 arcmin radius of any cen- tral coordinate contained therein. The base catalogue and the criteria for selection are discussed in Sect. 3. Most of these ATCA projects were intended to image various other LMC objects that were not PNe.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	This manual search identiﬁed projects that could pos- sibly contain PNe in images that might result in a PN RC detection. Each central coordinate and each frequency from all of the identiﬁed projects found in this way were manu- ally processed to create images from the data and then the images were examined at the coordinates of known (optical) PNe for RC emission detection.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Our project, ATCA C2367, which operated in 2010, was executed speciﬁcally to study RC PNe in the MCs. This project yielded most (26 out of 29) of the 3 cm and 6 cm detections in this study.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Seventeen ATCA projects were found to meet our selec- tion criteria as explained above. All of the raw data at our frequencies of interest from these projects were examined. Each project contained data from two to four frequencies. We manually processed all of the data to produce images at every frequency. Every image was examined for RC detec- tions at the PN coordinates. Due to diﬀerences in sensitivity (RMS noise level) and primary beam coverage from one im- age/frequency to another, PN RC detection at one frequency did not mean that detection at another frequency was to be expected. Only nine projects (see Table 1) had data that led to a PN RC detection. Four of those projects had PN RC detections at only one frequency.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Table 2 lists the projects that had images that did not contain PN detections and shows the pointing centres of each of those images. It also contains the RP catalogue references (see Sect. 3) to the PNe contained within approximately two thirds of the diameter of each full image since detections close to the edge of the images are not reliable. A total of 61 images were developed that did not contain any PN RC detections.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	The ATCA data were processed using miriad software (Sault et al. 2011). The miriad sigest task was used to de- termine the 1σ RMS noise level in a region essentially de- void of any strong point sources near each PN examined. For unresolved sources, the miriad imfit task was used to determine the integrated ﬂux density from the PN using a Gaussian-ﬁtting. For resolved sources, the pixels within a 1σ Jy beam−1 contour centred on the PN were used to esti- mate the PN integrated ﬂux density. A positive RC detection with either method was deﬁned as a ﬂux density estimate >3σ above the RMS noise.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Table 1. Radio data used in this study are identiﬁed with the ATCA project numbers. Data using the Compact Array Broad- band Backend (CABB) conﬁguration were acquired using the 2 GHz receiver bandwidth capability of the ATCA. Project C2908 is a MOSAIC survey of the LMC Supershell 4.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	referred to here as RP. The RP catalogue of the LMC was constructed from a portion of the Anglo-Australian Obser- vatory (AAO)/UKST Hα survey of the Southern Galactic Plane (Parker et al. 2005). The catalogue covers ≈ 25 deg2 centred on the LMC and contains 629 PNe found by using deep Hα and ‘Short Red’ images. Each PN was then veriﬁed using spectroscopic measurements and classiﬁed as ‘Known’, ‘True’, ‘Likely’, or ‘Possible’. This catalogue was further re- ﬁned with an extensive re-examination of the ’Likely’ and ’Possible’ categories using a multiwavelength analysis of the catalogue in Reid (2014). The subset of the RP catalogue used in our current study contains coordinates only from PNe classiﬁed as ‘Known’ or ‘True’.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	The RP catalogue numbers are written as ‘RPxxxx’ to stand for ‘[RP2006] xxxx’. LMC catalogue references from Sanduleak et al. (1978), which have the form ‘SMP LMC xx’, are referred to as ‘SMP Lxx’. For references to SMC PNe cat- alogued as ‘SMP SMC xx’, we use ‘SMP Sxx’ as a shorthand. In general, we use SMP catalogue numbers where possible, but for PNe which do not appear in the SMP catalogue, we use the RP catalogue numbers.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Table 2. ATCA projects that contained PN coordinates but did not yield PN RC detections. The listed coordinates represent the centre of the pointing.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	RP Catalogued PNe* This catalogue is a general inventory of LMC PNe known at time of publication and also contains PNe found by Henize (1956), Lindsay (1961), Lindsay & Mullan (1963), Lindsay (1963), Westerlund & Smith (1964), Sanduleak et al. (1978), Jacoby (1980), Morgan & Good (1992), Morgan (1994).
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	spectrum which may image diﬀerent parts of the PN from the RC data. The position of the central star (CS) is likely the best deﬁnition if it can be detected but RC imaging does not detect the CS.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	The synthesised telescope beam for each radio detec- tion is shown in the images of the PNe (Fig. 1–5). The syn- thesised beam is a two-dimensional Gaussian distribution dependent on the wavelength, the ATCA telescope array conﬁguration, and the coordinates of the image. The optical diameters of the PNe we studied here range over an order of magnitude – from 0.24 arcsec to 2.53 arcsec. The synthe- sized beam sizes are typically not symmetric and vary widely as seen in Table 1 (Col. 4) from 0.7 arcsec to 10.0 arcsec. In addition to meeting the requirement that the intensity measured is >3σ above the RMS noise, we require that a PN RC detection meets one of two additional conditions: 1) for unresolved sources the peak of the RC intensity must be <5 arcsec from the RP PN coordinates; or 2) the ap- parent centre of a resolved source must be <5 arcsec from the RP PN coordinates. We have chosen 5 arcsec since the maximum absolute positional uncertainty of the ATCA is 5 arcsec (Stevens et al. 2014).
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	After careful examination of all of our available data, 21 new RC detections of PNe in our base catalogue were iden- tiﬁed: SMP L13, SMP L15, SMP L21, SMP L23, SMP L29, SMP L52, SMP L45, SMP L37, SMP L73, SMP L63, SMP L53, SMP L75, SMP L76, SMP L78, RP 659, SMP L85, and SMP L92.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	For seven previously detected PNe (Filipovi´c et al. 2009): SMP L47, SMP L48, SMP L62, SMP L74, SMP L83, SMP L84, and SMP L89, a ﬂux density was measured on at least one frequency, conﬁrming previous detections. The data for these PN detections come from the various ATCA projects listed in Table 12. All of the RC LMC PNe de- tected here are from the set of PN coordinates that RP has published with a classiﬁcation of ‘Known’ or ‘True’.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	The four PNe listed in Filipovi´c et al. (2009) for which we did not ﬁnd any new reliable detections were SMP L8, SMP L25, SMP L33, and SMP L39. SMP L8 has been reclas- siﬁed as a compact H ii region and is not listed in the RP catalogue. Therefore, they are not included in our study pre- sented here. We detected RC emission from the positions of previously known RC PNe SMP L25 and SMP L33 in 20 cm data from ATCA project C354. Unfortunately, both PNe are far from the centre of the primary beam (∼25 arcmin at 20 cm) and therefore their ﬂux density measurements are very uncertain even after the image has been corrected for the primary beam (miriad linmos task). SMP L39 was not contained in any of the projects which met our requirements for inclusion.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Figure 1. New and previous radio detections of LMC PNe. Images are constructed as total intensity radio maps. Contours are integral multiples of the measured RMS noise starting at 3σ with spacings of 1σ. Top row: SMP L13 at 6 cm (grey scale) with 20 cm contours and SMP L15 at 6 cm (grey scale) with 20 cm contours; Second row: SMP L21 at 6 cm and SMP L23 at 13 cm (grey scale) with 20 cm contours; Third row: SMP L29 at 6 cm (grey scale) with 20 cm contours and SMP L37 at 6 cm. The beam size of each image is shown in the bottom left corner. The orange crosses represent the PN positions from the RP catalogue.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Figure 2. New and previous radio detections of LMC PNe. Images are constructed as total intensity radio maps. Contours are integral multiples of the measured RMS noise starting at 3σ with spacings of 1σ. Top row: SMP L38 at 6 cm (grey scale) with 20 cm contours and SMP L45 at 2.1 GHz; Second row: SMP L47 and SMP 48 at 13 cm (grey scale) with 20 cm contour; Third row: SMP L50 at 2.1 GHz and SMP L52 at 5 GHz. The beam size of each image is shown in the bottom left corner. The orange crosses represent the PN positions from the RP catalogue.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Figure 3. New and previous radio detections of LMC PNe. Images are constructed as total intensity radio maps. Contours are integral multiples of the measured RMS noise starting at 3σ with spacings of 1σ. Top row: SMP L53 at 5 GHz (grey scale) with 2.1 GHz contours and SMP L58 at 5.5 GHz; Second row: SMP L62 at 3 cm (grey scale) with 6 cm contours and SMP L63 at 9 GHz (grey scale) with 5 GHz contours; Third row: SMP L66 at 5 GHz and SMP L73 at 5 GHz. The beam size of each image is shown in the bottom left corner. The orange crosses represent the PN positions from the RP catalogue.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Figure 4. New and previous radio detections of LMC PNe. Images are constructed as total intensity radio maps. Contours are integral multiples of the measured RMS noise starting at 3σ with spacings of 1σ. Top row: SMP L74 at 3 cm (grey scale) with 6 cm contours and SMP L75 at 3 cm (grey scale) with 6 cm contours; Second row: SMP L76 at 5 GHz and SMP L78 at 5 GHz; Third row: RP 659 at 5 GHz and SMP L83 at 6 cm (grey scale) with 20 cm contours. The beam size of each image is shown in the bottom left corner. The orange crosses represent the PN positions from the RP catalogue.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Figure 5. New and previous radio detections of LMC PNe. Images are constructed as total intensity radio maps. Contours are integral multiples of the measured RMS noise starting at 3σ with spacings of 1σ. Top row: SMP L84 at 6 cm and SMP L85 at 5 GHz; Bottom row: SMP L89 at 3 cm (grey scale) with 6 cm contours and SMP L92 at 5 GHz. The beam size of each image is shown in the bottom left corner. The orange crosses represent the PN positions from the RP catalogue.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	are included in Table 3 with intensities from Filipovi´c et al. (2009) for completeness of the LMC PN sample. The ﬁnding charts for all LMC PN RC detections are shown in Figs. 1 through 5. These images are constructed as total intensity RC maps at the detection frequencies overlaid with contours at integral multiples of the estimated RMS noise. The orange crosses represent the PN coordinates from the RP catalogue. As can be seen from the ﬁnding charts presented (Fig. 1–5), in most cases the radio detections are of iso- lated sources well correlated with their RP catalogue optical counterparts. Most detections of a PN were at one or two frequencies with only two detections at three frequencies and none at four frequencies. The absence of detection at more frequencies was due either to a lack of data for a PN at that frequency or that the emissions were determined to be below the 3σ noise level.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Table 3. Coordinates and ﬂux density measurements of detected LMC PNe. The new ﬂux density measurements are shown in the columns titled ‘new’ and previous detections (Filipovi´c et al. 2009) are listed in the columns titled ‘F09’. The symbol ‘<’ indicates that the actual value is less than the indicated value. Col. 12 is the MIR/S20 cm ratio using 8 µm data from the VizieR Online Data Catalog: SAGE LMC and SMC IRAC Source Catalog (IPAC 2009). Col. 13 is the calculated Hα ﬂux using tabulated Hβ and extinction values cHβ from (Reid & Parker 2010). Col. 14 is from Table 4. The ‘Source Ref.’, Col 15, refers to the sources of the RC data used for our measurements from the ATCA project data base references found at the bottom of this table.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	ATCA project numbers as in Table 1 - 1: C256, 2: C308, 3: C354, 4: C395, 5: C479, 6: C520, 7: C1973, 8: C2367, 9: C2908.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	From the measurements made here and data from the litera- ture, additional physical properties of the detected PNe were determined. In Table 3 (Column 12) we calculated the ratio of Mid-Infrared (MIR) and 20 cm ﬂux densities using 8 µm MIR data from the Sage LMC and SMC IRAC Source Cata- log (IPAC 2009)3 containing LMC data from Meixner et al. (2006). This data was downloaded from the VizieR Online Data Catalog collection. Cohen et al. (2007b) using Galac- tic data from Cohen & Green (2001) and other sources refer- enced therein, examined the MIR-RC ratio, S8.3µm/S843 MHz for PNe in the Galactic plane and calculated a median ratio of 12. The median ratio for our population of the LMC PNe is 11.9 using S8µm/S1.4 GHz. The largest deviation from this median in our sample was calculated to be 87.9 for SMP L38. This, however, is likely to be due to unusual dust emissions from polycyclic aromatic hydrocarbons emitted by this PN as discussed in Bernard-Salas et al. (2009) who included this PN in their study of the MC PNe.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	where α is the spectral index and Sν is the ﬂux density at frequency ν. The spectral indices are shown in Cols. 4a and 4b of Table 4.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	PN RC emission is primarily thermal radiation from free-free interactions between electrons and ions within the nebular shell (Kwok 2000). The SED is dependent on the metallicity, electron density, and temperature. Typically, at higher frequencies (>5 GHz), the nebula is optically thin and the spectral index (α) is expected to be ∼ −0.1. At lower frequencies (<5 GHz), young PNe are most likely optically thick with an expected spectral index of ∼2 (Pottasch 1984). However, diﬀerent density proﬁle models can produce a very diﬀerent SED. For example, if the density proﬁle is from a constant mass loss, the spectral index for the PN would be optically thin and observed to be inverted with α ∼0.6 (Panagia & Felli 1975). Gruenwald & Aleman (2007) cau- tions that radio SED observations from PNe cannot uniquely determine the density proﬁles of the PNe.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	The critical frequency of a PN is deﬁned as the tran- sition frequency from optically thick with a positive spec- tral index to optically thin with a negative spectral index. Most Galactic PNe exhibit a critical frequency of less than 5 GHz (Gruenwald & Aleman 2007). However, the critical frequencies for the PNe depend on gas density and can range from ∼0.4 GHz to >15 GHz (Gruenwald & Aleman 2007). Young and compact PNe have been observed with critical frequencies of ∼5 GHz to 28 GHz (Aaquist & Kwok 1991). These young PNe most likely have very high electron den- sities which makes the critical frequency much higher than 5 GHz (Kwok 1982; Pazderska et al. 2009). We calculated the spectral indices α above and below 5 GHz from Eq. 1 as a proxy of the evolutionary state of the PNe.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Non-thermal radio emission has been proposed for some PN emissions (Dgani & Soker 1998). The mechanism oﬀered requires fast winds from the CS of the PN to create an inner region where strong magnetic ﬁelds are interacting with the CS wind. A non-thermal SED would be expected to have a spectral index α ≪ −0.1 (Gurzadyan 1997). Cohen et al. (2006) reported non-thermal emissions from a very unusual Asymptotic Giant Branch (AGB) star. This star was mea- sured to have α ∼ −0.9 and contained shocked interactions between a cold PN–like nebular shell and the hot, fast stel- lar wind from the AGB star. The source, however, of most apparent non-thermal radio detections is more likely to be a Supernova Remnant or an Active Galactic Nucleus coin- cident with the position of the PN.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	In Col. 5 (Table 4) the adopted ﬂux densities at 6 cm for all of the detected RC PNe are presented. If direct mea- surement at 6 cm was not available, we estimated the ﬂux density from neighbouring bands. This was necessary for SMP L23, SMP L33, SMP L39, and SMP L45. The entries that are estimated are indicated in Table 4 with brackets around the 6 cm values.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	We also compared our adopted 6 cm ﬂux densities with a sample of well-measured Galactic PNe (GPNe). This sam- ple is a data base of optical diameters and 6 cm inte- grated ﬂux density measurements from Si´odmiak & Tylenda (2001) consisting primarily of PNe from the Galactic bulge. Their diameter measurements were calculated using the 10 per cent peak contour approach. They used only optical data from photographic plates since they judged that newer CCD images were too sensitive and could not be directly com- pared to extant diameter measurements derived from pho- tographic plates. We re-scaled their data to the distance of the LMC (50 kpc; Pietrzy´nski et al. (2013)) and plot- ted both data-sets on the same ﬁgure (Fig. 6). Addition- ally we included nine SMC PNe from Leverenz et al. (2016): SMP S6, SMP S13, SMP S14, SMP S16, SMP S17, SMP S18, SMP S19, SMP S22, and SMP S24. These SMC PNe were unambiguously determined not to be PN mimics and were also re-scaled to the distance of the LMC. The green star markers on Fig. 6 represent the SMC data.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Our RC LMC PN measurements are consistent with the scaled ﬂux densities and diameters from both the Galactic and SMC samples. The angular diameters in Ta- ble 4 (Col. 2) were adopted from Shaw et al. (2001, 2006), Vassiliadis et al. (1998), and Stanghellini et al. (1999). For consistency, we chose only photometric radii deﬁned as the radius containing 85 per cent of the total emitted ﬂux (Shaw et al. 2001) since measurements based on this method were available from all of the sources we consulted for the radius data which we adopted. We estimated that the diﬀer- ences in the techniques for the diameter measurements for this comparison was not signiﬁcant (the 10 per cent peak ﬂux density used by Si´odmiak & Tylenda (2001) for the GPNe compared to the 85 per cent ﬂux density limit we employed for the MC PNe).
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Table 4. Measured and calculated LMC PN RC parameters. Col. 2 contains the photometric angular diameters. The references used for the diameters are listed in Col. 3 and they refer to the list at the bottom of the table. The spectral indexes (α) are listed in Col. 4a and 4b. The adopted ﬂux densities at 6 cm are in Col. 5. Those which are in brackets were estimated from other band measurements. The angular diameters in Col. 2 in arcseconds are used to calculate the radii in pc shown in Col. 7. Values for brightness temperatures (Tb), the electron densities (ne) and ionised masses (Mi) are calculated from the adopted 6 cm ﬂux densities and the radii. These are presented in Cols. 6, 8, and 9, respectively.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	parameter and the consensus ﬁlling factors were both as- signed a numerical value of 0.3 (Boﬃ & Stanghellini 1994; Frew et al. 2016a). The helium abundance for SMP L13 was taken from de Freitas Pacheco et al. (1993). SMP L39, SMP L52, and RP 659 were assigned the default value of 0.1 per Gathier (1987) since measured values for those PNe were not available. Calculations for the remaining PNe used he- lium abundances from Chiappini et al. (2009). No values for the fraction of doubly ionised helium were available so we adopted the default value of 1/3 as used by Gathier (1987). Uncertainties in ne and Mi are estimated to be on the order of 40 per cent.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Figure 6. Galactic PNe from (Si´odmiak & Tylenda 2001) (small red dots), SMC PNe data from (Leverenz et al. 2016) (green stars), (both scaled to the distance of the LMC), and our new LMC measurements (large black dots).
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	at 20 cm and 1.01 at 6 cm. The value of 339 K for Tb from SMP L33 is determined to have a correction factor of 1.4 at 20 cm. At 6 cm, there is essentially no self absorption. Our conclusion was that there was signiﬁcant self absorption in some of our results at 20 cm but at shorter wavelengths the self absorption was negligible.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	The radio ﬂux measurements at 6 cm and radius measure- ments in parsecs from Table 4 were used to calculate the surface brightness and diameter values for the Σ–D relation study of the MC PNe.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	the LMC (Reid & Parker 2013; Reid 2014) but only 31 of them have been detected at radio frequencies. This is indicative of a strong observational bias suggesting that only the ‘tip of the iceberg’ is observed at radio frequencies, see Vukoti´c et al. (2009). Indeed, while the measured surface brightness of GPNe extends from ∼ 10−24 to ∼ 10−16 Wm−2Hz−1sr−1 , the PNe from the MCs are detected only in the high brightness half of this interval from ∼ 10−20 to ∼ 10−16 Wm−2Hz−1sr−1. Similar selection eﬀects can be seen in D as well. The largest PN in our radio LMC sample is smaller than 1 pc, unlike the largest PN used for Hα statistical distance calibration (Frew et al. 2016a) being > 2 pc in size.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Table 5. Results of the Σ–D analysis at 6 cm. Col. 1 lists the sample designation with the number of data points in Col. 2 and the calculated correlation coeﬃcient in Col. 3. Cols. 4-7 list the values for the parameters of the ﬁtting from the Σ oﬀsets and their estimated uncertainties. The average fractional error for the calculated distances are in Col. 8. Cols. 9-13 present the values for the orthogonal- oﬀset-ﬁtting and the corresponding average distance fractional errors. In Cols. 14-16 we present the average fractional errors for the PDF distance estimators: mean, mode, and median.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	fects on our Σ–D samples. We utilised bootstrap statistics (Efron & Tibshirani 1993) to reconstruct the prior distri- bution of the Σ–D relation parameters. For each Σ–D data sample we applied both Σ oﬀset and orthogonal oﬀset analy- sis. In addition, we reconstructed the Σ–D probability den- sity function (PDF) using bootstrap based kernel density smoothing (for details see Bozzetto et al. 2017, and refer- ences therein). The reconstruction of the Σ–D PDF provides a better insight into observational biases and evolutionary features of each sample (see also Vukoti´c et al. 2014, and ref- erences therein). Unlike Vukoti´c et al. (2014), the PDF cal- culation is improved with the use of bootstrap based kernel density smoothing instead of bootstrap based centroid oﬀset smoothing. We applied a standard Gaussian kernel in order to calculate optimal smoothing bandwidths. For data in two dimensions a Gaussian product kernel was used. The opti- mal smoothing bandwidths (h) were selected by minimising the bootstrap integrated mean squared error (BIMSE) (for details see Bozzetto et al. 2017).
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	From the LMC and SMC data we deﬁned three samples: 1) the sample of 28 LMC PNe; 2) the sample of 9 SMC PNe with reliable ﬂux density estimates; 3) the sample consisting of the combined LMC + SMC PNe with 37 total PNe.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	As a measure of the sample spread and reliability we calculated the correlation coeﬃcient r and the average frac- tional error f . For the Σ–D statistical distances to the MCs, we calculated f and estimated the distances from the Σ–D relation (for a detailed description of statistical distance cal- culation see Vukoti´c et al. (2014)). The results are shown in Table 5 and Fig. 7.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	With the highest r value and the smallest values for f (Ta- ble 5) the SMC sample appears to be very compact and re- liable. However, the kernel density smoothing does not give convergent results for optimal smoothing bandwidths. This is caused by the small number of data points. The optimal smoothing bandwidths are calculated around ∼ 0.14 and ∼ 0.40 but repeated calculations do not give consistent re- sults.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	We note that the SMC sample consists of a small num- ber of data points that are diﬃcult to measure accurately in that they are very close to the survey sensitivity line. This suggests that the quality of this sample could be weaker than it would ﬁrst appear from a simple linear regression. Applying the bootstrap procedure could provide erroneous values for A and β and their uncertainties. The small num- ber of data points, even with the high correlation value, gives non-physical values for A and β when ﬁtting the re-sampled samples. This resampling with repetition is the essence of the bootstrap analysis (see Vukoti´c et al. (2014) and refer- ences therein).
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Instead of presenting these erroneous values, in Table 5 we show values from a simple empirical ﬁtting of a single line to the SMC sample without the application of the boot- strap analysis. Even though we do not consider these ﬁtted parameter values reliable we do utilise the SMC sample to form a more complete combined MC PN sample.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	where N is the number of data points in the sample, di is the measured distance to the object of the ith data point, and ds is a statistical distance to that object determined i either from the parameters of an empirical ﬁtting or from the PDF method. The PDF method for a ﬁxed Σ value gives a probability density over D. The values of D for the mean, mode, and median are used to calculate the corresponding statistical distances (Table 5).
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Figure 7. Σ–D data plots at 6 cm with the reconstructed data probability density distribution and orthogonal regression best-ﬁtting line: LMC (left) and LMC+SMC (right). The PDF contour plot is calculated on the 100 × 100 grid mapped on the plotted variables range. Contours levels are at 0.1, 0.5, 1.0, 1.5, 2.0 and 2.5. Optimal smoothing bandwidths are found at hlog D = 0.11 and hlog Σ = 0.30 for the LMC sample and hlog D = 0.09 and hlog Σ = 0.26 for the LMC+SMC sample.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	evolution. With the larger data spread, the LMC sample is certainly less sensitivity biased than the SMC sample but the somewhat large uncertainties, especially ∆β ∼ 0.4 − 0.5, indicate a possible reduced sample quality. Although less extreme than the SMC sample, the LMC sample bias could cause peculiar values for the ﬁtting of the parameters for the re-sampled sample.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Combining the SMC and LMC samples into one sample yielded the best results. The combined sample has smaller values for ∆A and ∆β compared to the LMC sample alone (Table 5). We also found that the average fractional dis- tance error calculated with the PDF method for the com- bined sample is f ∼ 16 per cent which is lower than that of the LMC sample alone which is f ∼ 18 per cent. This demonstrates the potential of RC MC PN samples to be used for statistical distance estimates. However, ∼ 18 per cent of the LMC’s 50 kpc distance is ∼9 kpc. Comparing this to a 1 kpc LMC depth (Pietrzy´nski et al. 2013), (Frew et al. 2016c) suggests that there is still signiﬁcant room for im- provement. This is likely to be achieved by more sensitive and more accurate radio surveys with larger samples.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	The value of β in the Σ–D relation is theoretically ex- pected to change from ∼1 to 3 over the lifetimes of the PNe (Uroˇsevi´c et al. 2009). Using a GPN sample they found an empirical slope of β ≃ 2.3 ± 0.4 by ﬁtting the Σ oﬀ- sets. The β values from Σ oﬀsets in this work, 2.3 for com- bined sample and 2.2 for the LMC sample alone, may im- ply a similar age for the MC PN population compared to their selected GPN sample. However, the more robust or- thogonal slope of 2.9 (Table 5) could give the impression that the MC PNe might be in a late stage of their evo- lution. Vukoti´c et al. (2009) argued that the MC PN sam- ples are biased with sensitivity selection eﬀects and conse- quently the empirical slopes should be steeper than that estimated from the Σ − D empirical data ﬁttings. Applying the bootstrap procedure to a sample of GPNe with reliable distances, Vukoti´c & Uroˇsevi´c (2012) obtained an orthogo- nal oﬀset β = 3.1 ± 0.4 which is consistent with our MC results presented in Table 5. Both the Σ oﬀset slope for the GPN sample from Uroˇsevi´c et al. (2009) and the orthogonal oﬀset slope from Vukoti´c & Uroˇsevi´c (2012) overlap within the estimated uncertainties of the slopes from Table 5.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	To estimate the accuracy of our results related to the Σ–D slope we simulated the inﬂuence of sensitivity related selec- tion eﬀects. This method is from Vukoti´c et al. (2009) who applied this technique to a very small sample of ﬁve PNe. We performed a similar simulation (although more advanced in some technical details) applying this method to our MC PN samples. This great improvement in the number of ra- dio MC PN detections should result in a signiﬁcantly more complete PN sample.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Figure 8. Our initial ﬁndings for the 6 cm data sample proba- bility density distributions of the orthogonal distances from the orthogonal oﬀset best-ﬁtting lines Lo for the LMC and the com- bined sample.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	having a diameter of Dmin = 0.058 pc (Table 4) and the fact that the largest PNe usually swell up to Dmax ∼ 2 pc, we distributed artiﬁcial points randomly generated in the [Dmin, Dmax] interval using the random number generator of Saito & Matsumoto (2008) with a uniform density on the log D scale equal to the density of the adopted 6 cm data. This resulted in a total of 55 artiﬁcial points (for the com- bined sample and 42 for the LMC sample) per generated sample in a given [log Dmin, log Dmax] interval. These points were then projected on to a series of lines with β ranging from 1.5 to 3.6. All of these lines intersected the orthogo- nal oﬀset best-ﬁtting line to the 6 cm adopted data (Lo) at Dmin. The next step was to disperse the points from the lines upon which they were projected in such a way as to simulate the dispersion of the adopted 6 cm data. This was achieved by applying the kernel density smoothing in 1D from Bozzetto et al. (2017). From the adopted 6 cm data we reconstructed the smooth probability density distribu- tion of the orthogonal distances (oﬀsets) from the Lo line (Fig. 8).
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	no signiﬁcant diﬀerence for the overall scatter in the case of the LMC and the combined sample. Both distributions were rather symmetric around the peak which was close to a value of zero. Each point was randomly dispersed in the orthogo- nal direction from the simulated slope line according to the probability density distribution from Fig. 8. The dispersion was controlled with a parameter called Scatter which is mul- tiplied by the dispersion obtained from the distributions in Fig. 8 to obtain the actual dispersion. We then calculated the orthogonal oﬀset ﬁtting parameters for the dispersed ar- tiﬁcial sample. For each simulated slope we generated 100 artiﬁcial samples. The mean values of the ﬁtting parameters after selection and their estimated uncertainties (calculated as the standard deviation of the given array of elements) are presented in Table 6. In Fig. 9 we plotted one artiﬁcial sam- ple for Scatter = 2.0 with a simulated β of 3.5 along with the combined MC PN sample and sensitivity line. Since the presented LMC survey is not uniform in terms of sensitivity we selected the sensitivity line in such a way that it passed through the data point of the combined sample that gives the lowest Σ value at the horizontal part of the sensitivity line. For the break point of the sensitivity line we selected the average equivalent of the circular beam size calculated from the beam sizes of each particular reduced image. At the adopted LMC distance this translates to D = 1.017 pc. The results in Table 6 do not appear to be signiﬁcantly dependent on Scatter. For each given Scatter the ∆β be- comes larger than σ<β> when the simulated β becomes as large as ∼ 2.5 − 2.7. This corresponds to < β >∼ 2.5 (Scatter = 1.0) and, although within the estimated uncer- tainty, reduces to < β >∼ 2.3 for a larger Scatter. Hence the steeper slopes, such as ∼ 2.9 (as obtained in this work) are signiﬁcantly inﬂuenced by the sensitivity selection ef- fect. For < β >∼ 2.9 the simulated slope is ∼ 3.4 and is even greater for larger Scatter. Since the MC PN samples appear to be signiﬁcantly inﬂuenced by sensitivity selection it is likely that only the brightest part of the sample is de- tected. However, any strong claims should be avoided since the results might depend on the simulation algorithm and the selected sensitivity line.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	We searched the ATOA for projects with high resolution data at 3, 6, 13, or 20 cm which had pointing centres within 10 arcmin of PN coordinates from the RP database. Some seventeen projects met those requirements and were processed in order to search for PNe at established (RP database) coordinates. We found 28 sources with detectable radio emissions in at least one of these radio bands with a ﬂux density ≥ 3σ above the noise. Of these 28 PNe, 21 are new radio detections reported here for the ﬁrst time. We also report a total of 61 ATCA images from the seventeen projects within which we did not ﬁnd any PN RC emissions at RP database coordinates that were ≥ 3σ above the noise. This radio-continuum tally is only ∼5 per cent of the known population of PNe in the LMC. The likely reasons for this very small detection ratio are the relatively low ﬂux densities from the PNe at the distance to the LMC and the lack of complete (and uniform) area coverage of the LMC with high sensitivity observations.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Figure 9. An example of an artiﬁcial sample generated from the combined sample orthogonal oﬀsets for the simulated β of 3.5 and a Scatter of 2.0. After selection the slope ﬂattens to 2.8 (short-dashed line ﬁt). The solid line is the sensitivity line with the horizontal part at Σ = 7.54 × 10−21Wm−2Hz−1sr−1 with the break point at D = 1.017 pc. The long-dashed line is the orthogonal oﬀset best-ﬁtting to the combined MC PN sample. Data points in black are the simulated sample with the points below the sensitivity line plotted as ﬁlled circles.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Our measurements along with data from the literature were used to calculate additional RC properties of the de- tected PNe. We calculated the MIR-20 cm ﬂux density ra- tios resulting in a median value of 11.9 compared to the Cohen et al. (2007a) value for the Galactic bulge of 12. We further calculated essentially no self absorption eﬀect for our 6 cm measurements. In addition, our adopted 6 cm ﬂux den- sities and diameters compare very well with the Galactic bulge distance-scaled PN data set from Si´odmiak & Tylenda (2001).
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Table 6. Parameters of the simulated samples and results of the sensitivity simulations for the LMC sample and the combined sample. Cols. 1-2 are the parameters from Equation 4 of the simulated Σ–D evolution lines. While Cols. 3-7 present the results for LMC sample, Cols. 8-12 in a same way present the results for the combined sample. In Cols. 3 and 5 are the mean values of Σ–D parameters calculated from the orthogonal oﬀset ﬁts after the selection has been applied and their estimated uncertainties are in Cols. 4 and 6, respectfully. In Col. 7 we give the diﬀerence between the mean slope after selection and the simulated slope. For all simulated slopes we ﬁt 100 artiﬁcially generated samples.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	liable data to examine the 6 cm radio Σ–D relation. Two sample sets were analysed: the LMC sample and the com- bined LMC+SMC sample. The reconstruction of the prob- ability density function shows that the rather complex PN evolutionary paths are not likely to be well represented by a single linear best-ﬁtting line but that more complex sta- tistical tools should be used. The rather incomplete SMC sample well complements the more numerous LMC sample. We ﬁtted the selected Σ–D data with both Σ oﬀsets and or- thogonal oﬀsets to obtain the ﬁtting parameters. The value of β in the Σ–D relation is theoretically expected to change from ∼1 to 3 over the lifetime of the PNe (Uroˇsevi´c et al. 2009). For the combined sample we obtained for orthogonal oﬀsets β = 2.9 ± 0.4 and for Σ oﬀsets β = 2.3 ± 0.3. Both values are comparable to the previously obtained values for GPN samples.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Sensitivity selection eﬀects were examined by simulat- ing an artiﬁcial random sample with values of β from 1.5 to 3.6 and values of D extending over a range from the min- imum observed diameter in our sample to 2 pc. A strong selection eﬀect is seen in MC data. The MC RC data is very close to the measurement sensitivity limit which weakens its quality and Σ is only detected in the high range of values expected from measurements taken of GPNe. The results of our Monte Carlo sensitivity simulation suggested that selec- tion eﬀects are signiﬁcant for values larger than β ∼ 2.6. We note that a measured slope of β = 2.9 should corre- spond to a sensitivity free value of ∼ 3.4. We demonstrated that an artiﬁcial sample derived from the combined sample shows that a large part of the data contained in the arti- ﬁcial sample cannot be detected with the sensitivity of the measurements in available RC data from the MCs.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	For the LMC and combined LMC+SMC samples, ap- plying the PDF statistical distance calculation, we calcu- lated the average fractional error for the statistical distance measurement. For the mean, mode, and median PDF esti- mators the average fractional distance error was ∼ 16 per cent for the combined sample and ∼ 18 per cent for the LMC sample alone. At the LMC distance (∼ 50 kpc) this ∼ 18 per cent translates to ∼ 9 kpc average distance error which is still signiﬁcantly larger than the 1 kpc LMC depth.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Our study demonstrated the potential for extended RC MC PN samples to be used as statistical distance estimators. Still, much can be improved, including the understanding of observational biases, especially sensitivity selection eﬀects. Next generation multi-frequency radio surveys with greater sensitivity and accuracy should lead to improved PN classi- ﬁcation and to a better understanding of PDF evolutionary features.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	ming language. DU and BV acknowledge ﬁnancial support from the Ministry of Education, Science and Technological Development of the Republic of Serbia through the project #176005 ‘Emission nebulae: structure and evolution’. We would like to thank the anonymous referee for the astute comments and careful reading of this paper that signiﬁcantly improved the ﬁnal version.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Henize K. G., 1956, ApJ Supplement Series, 2, 315 Jacoby G. H., 1980, ApJ Supplement Series, 42, 1 Kwok S., 1982, ApJ, 258, 280 Kwok S., 1985, ApJ, 290, 568 Kwok S., 2000, The Origin and Evolution of Planetary Nebulae.
Processed_Radio_Planetary_Nebulae_in_the_Large_Magellanic_Cl.txt	Panagia N., Felli M., 1975, A&A, 39, 1 Parker Q. A., et al., 2005, MNRAS, 362, 689 Pazderska B. M., et al., 2009, A&A, 498, 463 Pietrzy´nski G., et al., 2013, Nature, 495, 76 Pottasch S. R., 1984, Planetary nebulae. Astrophysics and space science library Vol. 107, D. Reidel ;Distributed in the U.S. by Kluwer Academic, Dordrecht ;Boston :Hingham, Mass.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	We use new and established methodologies in multivariate time series analysis to study the dynamics of 414 Australian hydrological stations’ streamﬂow. First, we analyze our collection of time series in the temporal domain, and compare the similarity in hydrological stations’ candidate trajectories. Then, we introduce a Whittle Likelihood-based optimization framework to study the collective similarity in periodic phenomena among our collection of stations. Having identiﬁed noteworthy similarity in the temporal and spectral domains, we introduce an algorithmic procedure to estimate a governing hydrological streamﬂow process across Australia. To determine the stability of such behaviours over time, we conclude by studying the evolution of the governing dynamics and underlying time series with time-varying applications of principal components analysis (PCA) and eigenspectrum analysis.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	With the recent focus on long-term shifts in temperature and weather patterns, and the potentially devastating impact for the future, the importance of earth and environmental science may never have been greater. There is a need for continued and varied studies of country hydrology behaviours, to ensure features such as streamﬂow are not exhibiting signiﬁcant changes in their behaviour. This paper explores streamﬂow patterns among Australian hydrology stations between 01 January 1980 and 01 January 2019.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	There is a signiﬁcant collection of work studying spatio-temporal patterns among hydrological processes [35, 36, 37]. Although the scope of such studies varies, there is generally a consistent theme of exploring the change in temporal behaviours at diﬀerent locations in space. One feature of particular interest to hydrology researchers has been streamﬂow modelling [38, 39], where a signiﬁcant amount of work has focused on modelling regional locations across the United States [40, 41, 42, 43, 44]. Furthermore, recent work has explored the use of wavelets in studying time series predictability within this domain [45]. Many of these studies have taken a more concentrated approach, providing detailed analysis on relatively few hydrology stations. This paper takes a diﬀerent approach, studying more than 400 stations, sampled from a broad range of geographic locations across Australia.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	A great deal of previous research in the hydrology community has used traditional linear and nonlinear statistical modelling techniques such as linear regression [46], multivariate linear regression and self-organizing maps (SOM) [47] and spatial modelling techniques [48] to yield insights related to topics such as El Niño Southern Oscillation (ENSO), drought, hydroclimatic variability and streamﬂow. In more recent years, many researchers have exploited advances in statistical learning, using various artiﬁcial intelligence methods such as support vector machines [49, 50] and deep learning [51, 52] to address problems related to prediction and assessment. These methods are often criticised for the diﬃculty one has in interpreting the drivers behind their predictions. In terms of speciﬁc applications of (relatively) recent work, some examples include Dickinson et al. and Najibi et al. [53, 54] who study the seasonality of climatic drivers behind ﬂood variability and drivers behind long-duration ﬂoods, respectively. Similarly Ryberg et al. [55] explore tree-ring based estimates of long-term seasonal precipitation, while Whitﬁeld et al. [56] explore spatial patterns of temporal changes in streamﬂow data. Our work asks similarly-themed questions and uses recently established and long-standing techniques in time series analysis to study temporal, spectral and spatial dynamics related to Australian streamﬂow, and subsequently explore collective similarity among our collection of hydrology stations.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	Although the study of nonstationary processes [57, 58], and the application of spectral analysis have been widely studied in the statistics community [59, 60, 61, 62, 63], there have been relatively few studies applying such techniques to address problems in hydrology. Notable studies of stationarity include the exploration of spatial correlations between daily streamﬂows [64], monthly streamﬂow forecasts [65], annual streamﬂow and ﬂood peaks [66]. Spectral analysis has been applied in hydrology to address several speciﬁc problems including the study of watershed properties [67] and aquiﬁers [68, 69, 70]. This paper uses a variety of techniques for the study of nonstationary processes and time-varying power spectra from the statistics literature, and applies them to model periodic phenomena in Australian streamﬂow. Furthermore, we explore the evolution of spectral patterns - testing for shifts in periodic dynamics over time.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	model the evolutionary dynamics of our underlying time series. Such studies of time-varying dynamics, correlation structure and dimensionality reduction have been widely applied among applied mathematicians in domains such as ﬁnance [71, 72, 73, 72, 74, 75, 76]. In hydrological research however, there is less abundance in the applications of such techniques [47]. We build upon studies of time-varying correlation structure to generate insights into collective temporal and spectral behaviours in Australian streamﬂow.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	Throughout this paper, we refer to a governing physical hydrological process. In this paper, we refer to the underlying dynamics governing Australian stream- ﬂow data evolving over space and time. In other hydrological contexts physical processes may refer to precipitation, energy balance, hydraulics, inﬁltration, etc. Throughout this paper, we focus on streamﬂow data.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	This paper is structured as follows. In Section 2 we describe the data used in this paper. In Section 3 we study the collective similarity of our underlying stations in the time domain. In Section 4 we apply an optimization framework and spectral analysis methodology to determine the similarity in various hydrological stations’ periodic behaviours. In Section 5 we introduce an algorithmic procedure to estimate a governing physical process, from the underlying stations. In Section 6 we explore the evolutionary dynamics in the governing physical process and the underlying time series, using a variety of techniques such as PCA and associated eigenspectrum analysis implemented in a time-varying capacity. In Section 7 we conclude.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	Our dataset consists of Australian hydrology stations, their location, catch- ment area and streamﬂow and is sourced from the Australian Bureau of Me- teorology (http://www.bom.gov.au/water/hrs). We only study stations with reported daily streamﬂow (ML/day) between 01-01-1980 - 01-01-2019. This provides a total of 414 stations to be studied throughout this paper.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	Figure 1: 4 candidate ﬂow time series from various locations in Australia. Each location clearly displays highly periodic behaviour.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	First, they all possess a similar annual spike in their streamﬂow values. Second, all locations display a similarly subdued level of streamﬂow prior to periodic spikes. Such periodic spikes appear to occur at similar times of the year, although this alignment between time series appears to deviate up to a small degree of perturbation. One key diﬀerence between the 4 stations visualised is the amplitude of their respective signals, which exhibit pronounced diﬀerences and can be seen by the upper bound of the y-axis. For instance, the peak daily streamﬂow in Paddy’s River at Riverlea is approximately 12,000 ML/day while the Gascoyne River at Fishy Pool is in excess of 800,000 ML/day.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	Figure 2: Temporal trajectory distance matrix applied to ﬂow time series. One can identify three predominant clusters (and a small outlier cluster) marked in purple, red, orange (and green) respectively. The red cluster displays a high degree of self-similarity relative to other clusters.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	Our resulting temporal aﬃnity matrix AT ∈ Rn×n is normalised such that all matrix elements now lie in [0, 1]. We apply hierarchical clustering to the resulting aﬃnity matrix, and study the collective temporal similarity.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	distinction between various inter-cluster and intra-cluster similarities. Four clusters of temporal streamﬂow time series are identiﬁed, three of which are most prominent. The two largest clusters, displayed at the top (purple) and bottom (orange) of the dendrogram both exhibit high levels of intra-cluster similarity, and although not quite as high, reasonably high aﬃnity between each other. The third primary cluster which is shown in red, displays markedly lower levels of similarity when compared with constituents of the purple and orange clusters. Having inspected the location of individual stations, cluster membership does not appear to be related to spatial location, but rather, temporal dynamics that are diﬃcult to detect.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	Given the pronounced similarity in the stations’ spikes in streamﬂow, we wish to further explore the periodic nature of this phenomenon. Spectral analysis is the most proliﬁc technique in the physical sciences for understanding a time series’ autocovariance structure. The most important properties to study are the peak amplitude and respective frequencies of a candidate power spectrum. In this section, we propose a Whittle [77, 78] Likelihood-based optimization framework, to learn the most appropriate parameters used in Welch’s method [79] for power spectral density estimation. In doing so, we are able to model the spectral dynamics of our entire collection of streamﬂow spectra. In describing our spectral analysis model for streamﬂow time series, we drop the subscript i, referencing speciﬁc time series, for notational convenience.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	We use Welch’s method to yield an estimate of the latent power spectral density. Welch’s method aims to reduce noise in its estimating the power spectrum by sacriﬁcing lower frequency resolution. Procedurally, the data is divided into overlapping segments where a modiﬁed periodogram is computed within each partition. The modiﬁed periodograms are averaged to produce a ﬁnal estimate of the power spectrum. Welch’s method consists of two model parameters: the length of each segment, and the amount of overlap in data points between adjacent segments. Overlap is often deﬁned in percentage terms, where 0% would mean that two adjacent segments have no data overlap while 50% would mean that any segment would share half the data of a neighbouring segment.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	Figure 3: 4 candidate ﬂow time series’ log power spectral densities from various locations in Australia. All power spectra appear to exhibit signiﬁcant power at low frequencies, with the power eroding across higher frequency components.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	Figure 4: Spectral trajectory distance matrix applied to ﬂow time series. There is one predominant cluster, and this highlights the overwhelming similarity in frequency behaviours among all streamﬂow time series. The similarity in periodic behaviour, relative to temporal behaviour, is striking.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	explore spectral similarity across the entire collection. We normalize for the amplitude of our power spectra with the same procedure as in Section 3. Let i (νj) = f ∗ i (νj ) ˜f ∗ i \|\| be our collection of optimal spectral trajectories. The vec- \|\|f ∗ tors ˜f ∗ i (νj) reﬂect relative diﬀerences in various stations’ spectra, and capture diﬀerences in key frequency components between streamﬂow spectra.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	and apply hierarchical clustering to the resulting spectral aﬃnity matrix. Again, this aﬃnity matrix AS ∈ Rn×n is normalised so that all elements lie in [0, 1]. A high score (close to 1) in the aﬃnity matrix AS would indicate that two hydrological stations exhibit similarly high power in key frequency components of their streamﬂow. A low score (close to 0) would indicate that two stations exhibit dissimilarity in their streamﬂow’s key frequency components.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	collection. There is one predominant cluster (consisting of a small sub-cluster), and an outlier cluster which consists of only a small subset of hydrology stations. This analysis conﬁrms that the periodic dynamics of the streamﬂow time series are highly uniform across the entire collection - where almost all streamﬂow spectra exhibit a high power low frequency component, and incrementally less power as one examines the spectrum against higher frequency components.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	Having observed the pronounced temporal and spectral similarity exhibited by Australian hydrological stations, regardless of their spatial location, we hypothesize that there may be a governing physical process which captures the dynamics of Australian streamﬂow. In this section we formulate an overall process, with each time series representing a potentially shifted version of this process.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	To minimize this loss function, we perform a sequence of updating steps. First, we update the governing process G. Here, we compute the mean of the current iteration of our governing process time series. Then, we update φ1, ..., φn. This step involves updating the alignments for the underlying stations’ time series. Then, we update the governing process, G, by computing another mean. Next, update φ1, ..., φn. This procedure converges quickly after a few iterations. (cid:80)n For initialization, we take G(t) = 1 i=1 xi(t) to be the overall mean of the n original, unshifted, time series.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	The initial estimate of the governing physical process and the estimate after the application of our algorithm are shown in Figure 5. The ﬁgure reveals several interesting takeaways. First, the spikes in aggregate aligned streamﬂow are highly periodic. First inspection suggests that a prominent spike in streamﬂow exists roughly once every year. This is certainly the most obvious pattern in the time series, with no clear insights in the periods between the big spikes. The peaks correspond to the signiﬁcant power observed at low frequencies in our spectral analysis.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	Figure 5: Temporal processes before and after our L2 alignment procedure. The ﬁgure shows the initial process prior to the alignment algorithm and the post-alignment procedure.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	process. An exhaustive list of non-zero oﬀsets, separated by state, are shown in Table 1. The three states with the most signiﬁcant number of non-zero oﬀsets are Tasmania, Victoria and Western Australia. Interestingly, these three states also share the largest median (non-zero) oﬀset value. However as a percentage of its total number of hydrology stations, Victoria’s 44 stations with non-zero oﬀsets is signiﬁcantly lower than Tasmania, Western Australia and Tasmania. As a percentage of the number of stations, the states that are most likely to exhibit phase shifting dynamics in the time domain are Tasmania (81%), Western Australia (69%) and South Australia (46.2%). As these states possess hydrology stations located at some of the most extreme points in Australia (by way of their spatial coordinates), this may indicate that there is some relationship between the oﬀset φ, and a candidate hydrology station’s spatial location.
Processed_Temporal_and_spectral_governing_dynamics_of_Austra.txt	Figure 6b shows the power spectral density for the governing physical process. Unsurprisingly, it displays the same key behaviours as the underlying stations’ streamﬂow time series. First, there is a low frequency high power component, indicative of the annual periodicity in streamﬂow, and reduced power in higher frequency components.

Subsets and Splits

No saved queries yet

Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.