THE GALACTIC POSITION DEPENDENCE OF FAST RADIO BURSTS AND THE DISCOVERY OF FRB011025

The High Time Resolution Universe (HTRU) survey, currently underway at the Parkes telescope (Keith et al. 2010), recently uncovered what appears to be an extragalactic population of single-bursting events (Thornton et al. 2013). Such events have previously been theorized to arise from a number of cataclysmic astrophysical events such as black hole evaporation (Rees 1977) and neutron star coalescence (Li & Paczyński 1998; Hansen & Lyutikov 2001). The four fast radio bursts (FRBs) published by Thornton et al. (2013) have so far represented the clearest evidence of such a population, followed by the additional recent discovery at the Arecibo Telescope by Spitler et al. (2014). If these FRBs are extragalactic, they hold vast potential as accurate probes of the ionized intergalactic medium (IGM), and for studying the exotic physical conditions that cause them. Unfortunately, the isolated millisecond-duration pulses detected by single-dish telescopes have poor sky localization, with positional errors of up to 30 arcmin.

Interferometric radio experiments will eventually provide localized FRB detections to precisely isolate burst origins. Meanwhile, any FRB discoveries will contribute to the distribution statistics of this potentially cosmological population. Understanding FRB distributions in dispersion, flux, and sky position are critical to make key assessments of likely FRB origins. Recently, the analysis of Petroff et al. (2014) indicated that there appears to be a dearth of detections in the HTRU survey intermediate latitude regions.

In this paper, we report a new FRB detection in archival Parkes Radio Telescope data and provide a robust method to assess the relative detection rates in a range of archival surveys at low Galactic latitude. We apply this method to a number of archival surveys to arrive at a confident statement regarding the Galactic latitude dependence of FRBs. Throughout this analysis, we take the following precautions with two previously purported FRBs:

• 1.  
  The Lorimer et al. (2007) detection. This burst's properties can be explained by both the terrestrial "Peryton" phenomenon of Burke-Spolaor et al. (2011a), and the FRBs of Thornton et al. (2013). Furthermore, this burst was discovered in a survey of the Magellanic Clouds, where there are significant errors in the Galactic electron density model used in our analysis (Cordes & Lazio 2002). To avoid introducing unwelcome biases from these ambiguities, we entirely exclude both the burst and the survey that gave rise to this detection from our analysis of Galactic position dependence of FRBs. This decision makes our study robust against future revelations that the Lorimer et al. detection is a member of either phenomenon.
• 2.  
  The Keane et al. (2012) detection. This burst's dispersion measure had only a small excess over that expected from a standard Galactic source, e.g., a neutron star. The report of Bannister & Madsen (2014) has shown, by analysis of Galactic ionized hydrogen in the Keane burst line of sight, that this burst has a high probability of being a Galactic object. We thus do not consider this burst in our analysis; however, we do employ its survey data.

We inspected the archival surveys of Edwards et al. (2001) and Jacoby et al. (2009), who collectively included Galactic latitudes 5° < |gb| < 30° at longitudes −100° < ℓ < 50°. The survey configurations were identical, sampling a 96-channel 288 MHz bandwidth centered at 1372.5 MHz. The time sampling for polarization-summed power samples was 125 μs, and each pointing lasted 4.4 minutes.

The original single pulse processing of these surveys was reported in Burke-Spolaor & Bailes (2010). We reprocessed the surveys using the search and automatic classification procedures of Burke-Spolaor et al. (2011b) which, in brief, involved seeking events over a signal-to-noise ratio (S/N) threshold of 6 in a range of dispersion measure and boxcar filter trials. Our boxcars were of size 2ⁿ, in the range 0 ⩽ n ⩽ 9. Candidate clusters are identified as events using a friend-of-friends method; single-member events are rejected as Gaussian noise, and events with a maximum S/N at DM ⩽ 1.5 pc cm⁻³ are automatically rejected as local interference. In addition to this process, to raise the sensitivity to bursts like those reported by Thornton et al. (2013), we:

• 1.  
  searched dispursion measure (DM) trials up to 3000 pc cm⁻³, higher than the previous search limit of 600 pc cm⁻³.
• 2.  
  employed pre-processing interference excision based on the coincidence filter techniques of Kocz et al. (2012) to minimize the false positive rate of burst candidates.

All events ranked as valid candidates were assessed manually using the same inspection plots and methods as Burke-Spolaor et al. (2011b). However, only pointing summary plots (not individual event plots) were inspected for pointings in which all candidates were of DM ⩽ 200 pc cm⁻³. As with other experiments of this kind, we estimate that candidates could be reliably identified down to S/N ≃ 7 (previous inspections of single pulses in a series of candidates from rotating radio transients have demonstrated that manual plot inspection is unreliable for S/N < 7; e.g., Burke-Spolaor et al. 2011b).

Only one burst with a DM exceeding the expected Galactic contribution, and not clearly associated with interference, was discovered in our search. The burst was detected at an S/N of 16.9. We report the observed properties of this burst in Table 1, and the data spectrogram is shown in Figure 1.

Zoom In Zoom Out Reset image size

This FRB has multiple properties that closely resemble the bursts reported by Thornton et al. (2013). First, the frequency-dependent delay closely follows a cold plasma dispersion relation, indicative of propagation through astrophysical plasmas (Figure 2), and unlike the terrestrial-origin "Perytons." The observed DM over that expected to be contributed by our Galaxy is a factor of ∼7 based on the ne2001 model's predicted dispersion at this line of sight, giving a large DM excess, DM_E, and indicative of an extragalactic origin for the burst. Assuming the fully ionized IGM model of Inoue (2004) and attributing all of DM_E to dispersion in the IGM, we obtain an upper limit on the redshift for our burst of z ⩽ 0.69. For comparison with the redshifts reported by Thornton et al. (2013), assuming a 100 pc cm⁻³ contribution from a galactic host indicates a redshift z = 0.59. This implies a maximum co-moving distance to the target of D ⩽ 2.4 Gpc, and D₁₀₀ = 2.1 Gpc, respectively.

Zoom In Zoom Out Reset image size

The burst has a duration of w₅₀ = 9.4 ms at its 50% power point. The width of the pulse is largely contributed by intra-channel dispersion smearing (7.9 ms; see Equation (2) below). Using Equation (2) to account for intra-channel smearing and the sample time, we find a 5 ms pulse width after removing instrumental broadening.

To test whether the pulse shows evidence of frequency-dependent scatter broadening, as the brightest of the Thornton et al. bursts did (they found w₅₀∝f^μ, where μ = −4.0 ± 0.4), we measured the pulse width in four independent sub-bands, and found that the pulse is broadened at low frequency with a best-fit index of μ = −4.2 ± 1.2, consistent with scatter broadening and with the index of the Thornton burst. Our pulse does not show significant asymmetry or a pulse tail (the light curve's skewness measured ±35 ms around the pulse peak is −0.018); however, noise in the data could easily be masking a scattering tail. The inferred pulse width of FRB011025 rivals that of the longest-duration FRB110220, which measured 5.6 ms and exhibited a clear scattering tail. The shorter durations of other FRB detections and our pulse broadening measurement imply that this pulse is likewise scatter-broadened.

Based on the similarity of their basic observed properties (excess dispersion, duration, fluence (time-integrated flux)), the burst appears to be from the same population as those reported by Thornton et al. (2013). We will assume that these bursts are all drawn from the same physical phenomenon until other evidence motivates a re-examination of this assumption.

A contrast in our discovery versus the Thornton et al. (2013) burst discoveries is the difference in the detected rate in our survey. Our archival intermediate-latitude data covered a total of 11858.76 hr on the sky, counting each 0.043 deg² beam in the multibeam receiver as an independent data stream. This implies a detection rate in the survey of ∼2 × 10³ sky⁻¹ day⁻¹, a factor of ∼5 smaller than the reported HTRU high-latitude detection rate. This gap widens when considering the non-detection of FRBs in other low-latitude surveys, notably the PKSMB survey of Manchester et al. (2001) and its Perseus Arm counterpart (Burgay et al. 2013), the PALFA survey reported by Deneva et al. (2009), and the search of 23.5% of the HTRU intermediate-latitude region, reported by Burke-Spolaor et al. (2011b). We aim to examine the statistical significance and origin of the detection deficit in these lower latitude surveys.

It is not immediately obvious whether this difference in detection rate is due to differences in sky coverage, or differences in sensitivity of each survey. Therefore, below we formulate the relative detection rates in surveys of different design and sky area.

4.1. Survey Design and Sky Position Considerations

For a pulse of intrinsic width t_i scattered to width τ_s, the minimum flux detectable by a survey performing a standard threshold-based search is given by

$$\begin{equation} S =\frac{m\eta (T_{\rm rec}+T_{\rm sky})}{G\sqrt{n_{\rm p}MBt_{\rm I}}} \times \frac{t_{\rm I}}{\sqrt{t_{\rm i}^2+{\tau _{\rm s}}^2}}, \end{equation} \tag{ 1 }$$

where m represents the S/N threshold for candidate pulses; η is an efficiency factor to account for sampling/correlator imperfections; G, n_p, and B give the telescope gain, number of polarizations, and the total bandwidth, respectively; and T_rec and T_sky represent the receiver and sky temperatures, respectively. M is a factor to account for dish configuration, e.g., M = 1 for single dish detection or M = n(n − 1) for imaging with an n-element interferometer. We assume that a system will integrate a pulse to time t_I ideally matched to the intrinsic width and scattering, within limitations of the native sampling time t_samp and pulse broadening effects due to finite channel width b:

$$\begin{equation} t_{\rm I}^2 = \left (\frac{t_{\rm i}}{\mu {\rm s}}\right)^2 + \left (\frac{t_{\rm samp}}{\mu {\rm s}}\right)^2 + \left (\frac{\tau _{\rm s}}{\mu {\rm s}}\right)^2 + \left (\frac{k{\rm DM}b}{f^3}\right)^2 \mu {\rm s}. \end{equation} \tag{ 2 }$$

Here, k = 8.3, b is in MHz, and f is in GHz.⁵ The scattering timescale is frequency-dependent; it may scale from a reference frequency such that τ_s = τ₀(f/f₀)^μ.

We now build a predictive expression based on Equations (1) and (2), and scaling factors for instantaneous field of view Ω, observation time t_obs, and spectral index. The number of detections expected in a single survey pointing p in survey 1, N₁(p), when compared to the total number of detections, N₂, discovered in the entire survey 2, is:

$$\begin{eqnarray} N_1(p) & =& N_2 \cdot \left (\frac{G_2m_1\eta _1\cdot [T_{\rm rec1}+T_{\rm sky1}(p)]}{G_1m_2\eta _2\cdot (T_{\rm rec2}+\langle T_{\rm sky2}\rangle)}\right)^\gamma \left (\frac{B_1M_1}{B_2M_2}\right)^{-\gamma /2} \nonumber\\ && \times \left (\frac{t_{\rm I1}(p)}{\langle t_{\rm I2}\rangle }\right)^{\gamma /2} \left (\frac{f_1}{f_2}\right)^{\alpha } \left (\frac{t_{{\rm obs}}(p)}{T_2}\right) \left (\frac{\Omega _1}{\Omega _2}\right) ; \nonumber\\ &=& N_2\cdot \theta (p). \end{eqnarray} \tag{ 3 }$$

Here, the index 1 or 2 represents the respective survey, T_sky is the sky temperature in the pointing direction of the observation, and T₂ is the total observing time of survey 2. The exponent γ is a power-law index representing the observed flux distribution of the population (N∝S^γ). The use of a power-law index here implicitly assumes that we can model any redshift-dependent evolution of the FRB population's luminosity with a power law. For a non-evolving population in a Euclidean universe, γ = −3/2; this is a reasonable assumption for FRBs for a generic z ≲ 1 astrophysical population. The averages over T_sky2 and t_I2 represent the pointing-averaged values for that survey; in the calculation of t_I2, the average values of DM and τ_s across the survey positions may be used if accounting for DM and τ_s effects. We define the spectral index as S∝f^α. T_sky is also frequency-dependent (it can change by a factor of two across a ∼300 MHz band at 1 GHz); however, we use the center frequency of each survey in this analysis to represent the pointing's value.

θ(p) encompasses all the expected scaling factors for pointing p. In our analysis below, we compare the detection rate in the archival surveys with the detection rate of Thornton et al. (2013), such that "survey 2" uses the parameters of Thornton et al., and N₂ = N_T = 4. The total number of events expected in the archival surveys is the sum of Equation (3) over every pointing in each survey:

$$\begin{equation} N_{\rm A} = N_{\rm T}\cdot \sum ^{n_{\rm surv}}_{s=1} \sum ^{n_{\rm point}}_{p=1} \theta _s(p) = N_{\rm T}\Theta, \end{equation} \tag{ 4 }$$

where we have simplified the additive scaling for all pointings in our survey set to be represented by Θ.

4.2. Inputs for "Local" and "Cosmological" Populations

We can now use the sky-dependent scalings of Equation (3) to demonstrate the anticipated detection rate versus sky position for FRBs. With an extragalactic population, one must account for gl and gb-dependent contributions to DM, τ_s, and T_sky, which may serve to dampen the number of expected detections at decreasing |gb|. However, any signal not propagating through the Galaxy (i.e., bursts of instrumental, terrestrial, or even solar system origin)—hereafter we refer to these as a "local/isotropic" population—will have detection rate dependencies affected only by T_sky, which effectively raises the receiver noise at low |gb|. For the remaining analysis in this paper, we assume basic properties of the FRB population based on what has thus far been observed, using the four bursts of Thornton et al. (2013) and our burst.

4.2.1. Local/Isotropic Population

Here we refer to any terrestrial or solar system origin which has no intrinsic clustering with respect to Galactic position. For instance: local FRBs might arise from self-generated signals in the receiver or software path, or from an atmospheric origin (which might not be isotropic in azimuth/elevation, but which would be uncorrelated with Galactic coordinates). Aircraft or satellite origins may appear isotropic, depending on the timing and route of the flight path. Likewise, "local" signals from planetary origins in or out of the ecliptic plane should appear isotropically distributed with regard to Galactic position, as long as the survey being searched is well-distributed in observing time and day of year.

A detection rate dependence on sky position for a local FRB population will arise only from the scaling (T_rec + T_sky)^γ. All other factors in Equation (3) will not induce position dependence for a local FRB population. We compute the position-dependent T_sky at 1.4 GHz using the multi-frequency global sky temperature model of de Oliveira-Costa et al. (2008). The resulting dependence of detection rate with sky position for a local population is shown in Figure 3 for a range of γ values.

Zoom In Zoom Out Reset image size

4.2.2. Cosmological Population

For the extragalactic FRB scenario, in addition to sky temperature, the factors that will influence the position dependence of the FRB detection rate are excess scattering and dispersion induced by the Milky Way's interstellar medium. These terms enter the rate through Equation (2). We take the mean extragalactic contribution to DM for the population to be 〈DM_E〉 = 772 pc cm⁻³. To evaluate the DM at which FRBs will be detected, for each Galactic position we add 〈DM_E〉 to the total Milky Way DM contribution at that position, DM_MW, employing the current standard ne2001 electron density model of the Galaxy (Cordes & Lazio 2002) to evaluate DM_MW.

We approximate Galaxy-induced scattering (τ_s) by using the ne2001 model to determine the distance, d_sc, at which the Galactic scattering timescale is maximized (τ_max). That is, τ_max is the largest scattering that can be induced in any pulse along that line of sight according to the ne2001 model; for extragalactic-origin FRBs, the scattering will always be smaller than this value.

To estimate the Galactic scattering experienced by an extragalactic FRB, we follow the analysis of Lorimer et al. (2013), taking τ_s = 4τ_max(1 − f)f, where f is d_sc divided by the source distance. We thus are required to select an average detectable FRB distance with which to estimate the magnitude of Galactic scattering. As long as we select a sufficiently high value, it will not greatly impact the results of this analysis because for current instrumentation, only FRB distances below a few hundred Mpc will cause Galactic scattering to dominate Equation (2) for Galactic latitudes beyond |gb| ∼ 3. FRB distances only have upper limits, so we use 1 Gpc to represent an arbitrarily large average detectable FRB distance, such that f = d_sc/(1 Gpc). We make the assumption that all bursts have an intrinsic duration t_i much less than instrumental broadening and scattering effects, and use μ = −4.0 to appropriately scale Galactic scattering (Bhat et al. 2004). The results of this analysis for two instruments are shown in Figure 4.

Zoom In Zoom Out Reset image size

4.3. Relative Detection Rates in the High-latitude HTRU and Low-latitude Archival Surveys

We now compare the detections of low-latitude archival transient searches with their expected FRB discovery rate based on the high-latitude detections of Thornton et al. (2013), using the analysis in Section 4.2 to examine expectations for the local/isotropic and extragalactic sky distribution hypotheses.

We calculated N_A using Equation (4) with the average properties of the Thornton et al. (2013) survey (i.e., DM₂ = 〈DM_MW〉 + 770 pc cm⁻³ = 819 pc cm⁻³; 〈T_sky2〉 = 0.85 K; 〈τ_s2〉 = 36.2 μs, as reported in Table 2). We report N_T∑_pθ(p) for each survey and the net N_A value in the bottom rows of Table 2, using local and extragalactic factors to calculate θ(p). Figure 5 demonstrates the N_A expected for a range of γ values.

Zoom In Zoom Out Reset image size

Table 2. Survey Parameters of Various Transient Searches and Results of Our Relative Detection Rate Analysis

	Parkes Archival Surveys			HTRU		PALFA
	SWMB	PKSMB	Perseus Arm	T13	P14	D09	S14
gb range	5° < |gb| < 30°	|gb| < 5°	|gb| < 5°	All-sky,	|gb| < 15°	|gb| ⩽ 5°	|gb| ⩽ 5°
gl range	−100° → 50°	−100° → 50°	200° → 260°	|δ| > 10°	−120° → 30°	30° → 78°;	162° → 214°
						162° → 214°
Trec	23			23		30
η	1.5			1.07		1.67
G (Jy K−1)	0.64			0.64		8.514
m	7	7	7	10	10	7	7
Ω (deg2)a	0.043			0.043		0.0027
f (MHz)	1372.5	1372.5	1374	1352		1440
B (MHz)	288			340		100
b (MHz)	3			0.390625		0.390625
tsamp (μs)	125	250	125	64		64
Tobs (h) a	11,859	20,248	6,924	7,849	15,041	3,214	1,974
DM limit	3000	5000	1500	2000	5000	1000	2038
DMMW $\bar{x}, \tilde{x}$	160, 120	701, 600	232, 193	49, 36	383, 289	396, 190	180, 182
$T_{\rm sky}\ \bar{x}, \tilde{x}$ (K)	1.70, 1.43	6.14, 5,14	1.21, 1.19	0.85, 0.70	3.18, 2.22	3.35, 1.84	1.31, 1.31
$\tau _{\rm s}\ \bar{x}, \tilde{x}$ (μs)	188, 0.18	11.7e3, 70.6	4.75, 3.06	36.2, 4.75e–3	4125, 2.78	1.31e3, 2.79	1.79, 1.74
[γ = −1.5]								TOTAL
NA, local	1.17	1.61	0.71	(4)b	6.81	1.19	0.81	12.3
NA, exgal	1.08	1.07	0.61	(4)b	5.46	0.90	0.73	9.85
NA, actual	1	0	0	(4)b	0	0	1	2

Notes. Surveys are: our search (Swinburne Multibeam surveys, SWMB), the Parkes Multibeam Survey (PKSMB; Keane et al. 2010, 2011; Mickaliger et al., in preparation), the Perseus Arm survey (Burgay et al. 2013; Thornton et al. 2013, hereafter T13), the HTRU intermediate latitude survey (Petroff et al. 2014, hereafter P14; see also Keith et al. 2010; Burke-Spolaor et al. 2011b), and two PALFA surveys, which used the 7 beam 21 cm transient search system on Arecibo (Deneva et al. 2009, hereafter D09 and Spitler et al. 2014, hereafter S14). For the SWMB, T13, D09, and S14 surveys, we computed DM_MW, T_sky, τ_s, and the numerical results using actual survey pointing lists (T13 pointings; D. Thornton & B. Stappers 2013, private communication; D09; J. Deneva & A. Brazier 2013, private communication). For the other surveys, where pointing lists were unavailable, we calculated these values using an evenly spaced grid of pointings across the quoted sky area of the survey. ^aNote that we have quoted, and used in computations, the single-beam field of view for the 7 and 13 beam PALFA and Parkes receivers; the total observation time reflects this accordingly. Numerical results are given for γ = −1.5 (see also Figure 5). ^bThese values were used as reference values (Equation (3)) to make the prediction values for the other surveys, and therefore are not included in the total sum.

Download table as:  ASCIITypeset image

Two bursts have been detected in the archival surveys we analyze. To understand the statistical significance of the difference between this and the expectations reported in Table 2, we take the probability of getting N_A(= 2) events given that N_T(= 4), which has a Poissonian distribution.

First we test the local/isotropic distribution hypothesis. As seen in Figure 5, the prediction is closest to our detected value at its minimum at γ ≃ −1.6. At that point, we reject a uniform sky distribution at a probability of P = 0.00037, which corresponds to the equivalent of a confidence of ∼3.6σ. In other words, even after accounting for the known effects of T_sky and instrumentation sensitivity, the observed burst rate's sky distribution is non-uniform to high significance, with a deficit at low galactic latitudes, for any value of γ.

In our extragalactic origin hypothesis, accounting for ne2001 predictions for scattering and dispersion at low Galactic latitudes helps to reconcile the low-latitude deficit; however, it still does not provide a confident agreement with the archival surveys' singular discovery. With our extragalactic formulation, N_A is at a minimum at γ = −1.7. The probability of detecting one burst at that γ is P = 0.0031, while the probability at the standard Euclidean γ = −1.5 is P = 0.0025, reflecting a disagreement at a confidence of around 3σ. This is not sufficient to rule out that FRBs are extragalactic, particularly considering our simplifying assumptions about the Galaxy and the FRB population. To resolve the discrepancy here, the extragalactic sky-distribution model would have to account for even fewer bursts detected in the low-latitude archival surveys. We review our assumptions here, along with how they might change the predicted numbers, and a note on what might account for the observed discrepancy:

• 1.  
  The average detectable FRB distance could be much closer than 1 Gpc. As previously noted, a very nearby FRB population would imply heightened Galactic scattering, decreasing the predicted number of detections at intermediate Galactic latitudes. We have investigated this possibility, and find that FRBs would have to be at an average distance of <100 Mpc to change the N_A predictions in the extragalactic model by more than 5%.
• 2.  
  We used the ne2001 model to predict Galactic scattering and dispersion effects. Additional Galactic scattering or other pulse flux dampening that is not correctly modeled by ne2001 may be occurring in the Galactic plane or halo. If such effects have a large gradient across gl and gb, they would have a major impact on our predictions. ne2001 is currently seen as the most accurate Galactic model; however, its errors increase at larger |gb|. Subsequent publications have argued that the scale height of the "thick disk" in ne2001 should be larger (Gaensler et al. 2008; Schnitzeler 2012). Such an alteration of the Galaxy model would raise DM_MW at intermediate latitudes, and significantly ease the observed discrepancy for the extragalactic population predictions in our analysis.
• 3.  
  We assumed that FRB pulses are distributed according to a power-law flux distribution (N∝S^γ). If FRB distributions are non-Euclidean (i.e., they are truly cosmological, and/or the progenitor population evolves significantly at z < 1), this may account for the fewer detections at the lower Galactic latitudes, where large T_sky contribution curtails sensitivity to faint pulses.
• 4.  
  We have assumed that typical FRBs have negligible intrinsic pulse widths and extragalactic scattering when compared to instrumental effects. Non-negligible values would actually increase the expected number of discoveries in the archival surveys, and therefore would not resolve the observed differences.
• 5.  
  Our assumption that all surveys have been searched to a sufficient DM may not be valid. We can assess this by inspecting the mean/median DM_MW values for each survey in Table 2, when compared to the DM search limit and considering the range of FRB DMs detected thus far. We see that this is a valid assumption for all but the Deneva et al. (2009) search, for which two of five of the known FRBs might not have been spotted in the search if they occurred in the most central galaxy regions. One additional discovery from this search at higher DM would ease, but not solve, the discrepancy with our distribution predictions. The rejection of a local population would still hold at a confidence of over 3σ.
• 6.  
  If the dispersion of the Keane et al. (2010) detection is not in fact due to the excess Galactic electrons reported by Bannister & Madsen (2014), this helps to bring the discovered and predicted populations closer. While this is true for both extragalactic and local predictions, the extragalactic prediction still maintains a better fit to the discoveries to date.
• 7.  
  We have assumed a flat spectral index at 1 GHz. However, this will not greatly affect our results as all surveys we analyzed were observed at similar frequencies.

Finally, one remaining possibility to explain disagreements with both of our prediction scenarios is that FRBs have a Galactic origin. However, there are strong observational arguments against this possibility (Kulkarni et al. 2014), and a Galactic explanation for FRBs must account for a lower detection rate in the Galactic plane; even very nearby Galactic sources show tendencies toward Galactic-plane excesses.

We end on one final note. We are clearly dealing with small number statistics (i.e., 1 and 4). Although our statistical statements account for Poissonian variance, they do not account for human error; Figure 5 demonstrates that if one burst was missed per low-latitude survey, we might see complete agreement with the extragalactic model predictions, and less significance in the disagreement with a uniform sky distribution or local population. We believe we have sufficiently accounted for this issue by restricting survey thresholds to S/N > 7, the point at which individual pulsar pulses are reliably flagged by human inspectors in manual inspection plots (to a ∼99% success rate). It is furthermore possible to miss FRB events due to radio interference that can render an entire pointing unusable. However, the occurrence rate of this effect is consistent between surveys and so cannot cause the variance with Galactic position that we have demonstrated here.

We have presented the discovery of a |gb| = 20° FRB, FRB011025, in a search of the intermediate-Galactic-latitude surveys of Edwards et al. (2001) and Jacoby et al. (2009). As evidenced by its flux, few-millisecond timescale, temporal isolation, a frequency sweep that closely adheres to the cold plasma dispersion relation, and its ∼7 times excess DM over that expected from the Galaxy, the burst appears to be from the same population as the bursts reported by Thornton et al. (2013). However, the detection rate in our search is a factor of ∼5 less than that of Thornton et al. (2013).

This difference led us to inspect the reason for the lack of FRB discoveries in a number of low-latitude f ≃ 1 GHz surveys as a means to test the sky distribution of FRBs against expectations for local and extragalactic FRB origins. Based on the survey design and discoveries of Thornton et al. (2013), we made predictions for the number of FRBs that should be discovered in the low-latitude archival surveys. Instrumental, sky temperature, and other Galactic filtering effects were considered (Equation (3)).

A test of FRBs as a local/solar-system origin population (i.e., uniform sky distribution with respect to Galactic position) revealed a rejection of this hypothesis at a confidence of 3.6σ. That is, we see a significant drop in the detection rate at low Galactic latitudes, which is strong evidence of a non-local (Galactic or extragalactic) origin.

A test of FRBs as an extragalactic population, using the ne2001 model to account for excess dispersion and scattering in the Galaxy, showed a closer agreement with discovery rates; however, FRBs still demonstrated a discrepancy at the ∼3σ level. An inspection of our simplifying assumptions revealed several points that could produce fewer expected detections at |gb| < 30°, and thus give a closer agreement with an extragalactic FRB model: most prominently, (1) an FRB population at <100 Mpc distance, and (2) excessive dispersion or scattering effects in the Galaxy/halo that are not accounted for by the ne2001 model, e.g., a larger disk scale height.

We have not ruled out the possibility of a Galactic origin for FRBs, which could produce more complex sky-dependent detection effects than were considered here. Kulkarni et al. (2014) provides an overview of several potential Galactic and extragalactic FRB progenitor models, and a look at the credibility of these possibilities.

These results must be interpreted with some caution: the small total number of FRB detections thus far, and their apparent rarity, make numerical prediction comparisons sensitive to human error during the manual inspection stage of candidate identification and other mis-estimations of sensitivity thresholds. However, we believe we have suitably accounted for these effects in our analysis.

Ongoing and future surveys, particularly those which focus on Galactic latitudes |b| < 15°, will place strong constraints on the latitude dependence of FRBs. Based on our framework, we expect 3.8 and 6.8 discoveries in the HTRU low latitude survey for extragalactic and local populations, respectively. For a survey of the inner Galaxy with the PALFA system, 1.5 and 2.8 bursts are expected from extragalactic and local populations, respectively, for a 1000 hr survey. However, the predictions for extragalactic event rates are likely to actually be lower given our results and when considering the simplifying assumptions made in our predictions (summarized in Section 4.3).

Finally, we would like to make the general note that regardless of FRB origins, future searches for FRBs will apparently find greater success if they survey at Galactic latitudes |gb| ≳ 20. The expression given in Equation (3) provides a convenient estimator for various survey design parameters to optimize FRB detection.

The Australia Telescope Compact Array Parkes Radio Telescope is part of the Australia Telescope National Facility, which is funded by the Commonwealth of Australia for operation as a National Facility managed by CSIRO. Processing for our archival search was performed on the Swinburne University Green Machine; we acknowledge the use of this supercomputer facility in this work. We thank B. Stappers, D. Thornton, J. Deneva, and A. Brazier for providing information about survey pointing directions which were used in this paper's analysis. We thank an anonymous referee for contributing valuable comments on this paper. Cosmological redshift calculations used the online calculator provided by Wright (2006).