Truncated Models for Probabilistic Weighted Retrieval

Existing probabilistic retrieval models do not restrict the domain of the random variables that they deal with. In this article, we show that the upper bound of the normalized term frequency ( tf ) from the relevant documents is much smaller than the upper bound of the normalized tf from the whole collection. As a result, the existing models suffer from two major problems: (i) the domain mismatch causes data modeling error, (ii) since the outliers have very large magnitude and the retrieval models follow tf hypothesis, the combination of these two factors tends to overestimate the relevance score. In an attempt to address these problems, we propose novel weighted probabilistic models based on truncated distributions. We evaluate our models on a set of large document collections. Significant performance improvement over six existing probabilistic models is demonstrated.

Modeling of Before Closure Zero Slope Pressure Derivative in a Diagnostic Fracture Injection Test DFIT

Recent diagnostic fracture injection test (DFIT) data presented on a Bourdet log-log diagnostic plot showed derivative slope of 0 in the before closure (BC) portion of the DFIT response. Some works qualitatively describe it as radial flow. This behavior has not been quantitatively analyzed, modeled and matched. The present work disagrees with the hypothesis of radial flow and successfully matches the relatively flat trend in the Bourdet derivative with a model dominated by friction dissipation coupled with tip extension. The flat trend in Bourdet derivative occurs shortly after shut-in during the before closure period. Because a flat derivative trend suggests diffusive radial flow, our first approach was to consider the possibility that an open crack at a layer interface stopped the fracture propagation and caused the apparent radial flow behavior observed in falloff data. However, a model that coupled pressure falloff from diffusive flow into a layer interface crack with pressure falloff from closure of a fracture that propagated up to the layer interface failed to reproduce the observed response. Subsequently, we discovered that existing models could match the data without considering the layer interface crack. We found that data processing is very important to what is observed in derivative trends and can mislead the behavior diagnosis. We succeeded to match one field DFIT case showing an obvious early flat trend. The presence and dominance of geomechanics, coupled with diffusive flow, disqualify the description of the flat trend in Bourdet derivative as radial flow. Instead, flow friction coupled with tip extension can completely match the observed behavior. Based on our model, cases with a long flat trend have large magnitude near-wellbore tortuosity friction loss and relatively long tip extension distance. Further, we match the near wellbore tortuosity behavior with rate raised to a power lower than the usually assumed 0.5. The significance of these analyses relates to two key factors. First, large magnitude near wellbore tortuosity friction loss increases the pressure required for fracture propagation during pumping. Second, tip extension is a way to dissipate high pumping pressure when very low formation permeability impedes leakoff. Matching transient behavior subject to the presence of both of these factors requires lowering the near-wellbore tortuosity exponent.

No fingers, no SNARC? Spatial-Numerical Associations and temporal stability of finger counting

The Spatial-Numerical Association of Response Codes (SNARC) effect (i.e., faster left/right side responses to small/large magnitude numbers, respectively) is considered as strong evidence for the link between numbers and space. The studies have shown considerable variation in this effect. Among the factors determining individual differences in the SNARC effect is the hand an individual uses to start the finger counting sequence. Left-starters show a stronger and less variable SNARC effect than right-starters. This observation has been used as an argument for the embodied nature of the SNARC effect. For this to be the case, one must assume that the finger counting sequence (especially the starting hand) is stable over time. Subsequent studies challenged the view that the SNARC differs depending on the finger counting starting hand. At the same time, it has been pointed out that the temporal stability of finger counting starting hand should not be taken for granted. Thus, in this preregistered study, we aimed to replicate the difference in the SNARC between left- and right-starters and explore the relationship between the temporal stability of finger counting starting hand and the SNARC effect. We expected that higher stability should be associated with a stronger SNARC effect. Results of the preregistered analysis did not show the difference between left- and right-starters. However, further exploratory analysis provided weak evidence that this might be the case. Lastly, we found no evidence for the relationship between finger counting starting hand stability and the SNARC effect. Overall, these results challenge the view on the embodied nature of the SNARC effect.

Large silicic magma bodies and very large magnitude explosive eruptions

Over the last 20 years, new concepts have emerged into understanding the processes that lead to build up to large silicic explosive eruptions based on integration of geophysical, geochemical, petrological, geochronological and dynamical modelling. Silicic melts are generated within magma systems extending throughout the crust by segregation from mushy zones. Segregated melt layers become unstable and can assemble into ephemeral upper crustal magma chambers rapidly prior to eruption. In the next 10 years, we can expect major advances in dynamical models as well as in analytical and geophysical methods, which need to be underpinned in field research.

Accommodating ~9 m of dextral slip on the Kekerengu Fault through ground deformation during the Mw 7.8 Kaikōura Earthquake, November 2016

<p>The Mw 7.8 Kaikōura earthquake of November 14th 2016 provided unprecedented opportunities to understand how the ground deforms during large magnitude strike-slip earthquakes. The re-excavation and extension of both halves of a displaced paleoseismic trench following this earthquake provided an opportunity to test, refine, and extend back in time the known late Holocene chronology of surface rupturing earthquakes on the Kekerengu Fault. As part of this thesis, 28 organic-bearing samples were collected from a suite of new paleoseismic trenches. Six of these samples were added to the preferred age model from Little et al. (2018); this updated age model is now based on 16 total samples. Including the 2016 earthquake, six surface rupturing earthquakes since ~2000 cal. B.P. are now identified and dated on the Kekerengu Fault. Based on the latest five events (E0 to E4), this analysis yields an updated mean recurrence interval estimate for the Kekerengu Fault of 375 ± 32 yrs (1σ) since ~1650 cal. B.P. The older, sixth event (E5) is not included in the preferred model, as it may not have directly preceded E4; however, if this additional event is incorporated into an alternative age model that embraces all six identified events, the mean recurrence interval estimate (considered a maximum) calculated is 433 ± 22 yrs (1σ) since ~2000 cal. B.P. Comparison of structures on an identical trench wall logged both before and after the 2016 earthquake, and analysis of pre- and post-earthquake high resolution imagery and Digital Surface Models (DSMs), has allowed the quantification of where and how ~9 m of dextral-oblique slip was accommodated at this site during the earthquake. In addition to this, I analyse the coseismic structure of the adjoining segment of the 2016 ground rupture using detailed post-earthquake aerial orthophotography, to further investigate how geological surface structures (bulged-up moletrack structures) accommodated slip in the rupture zone. These combined analyses allowed me to identify two primary deformation mechanisms that accommodated the large coseismic slip of this earthquake, and the incremental effect of that slip on the structural geology of the rupture zone. These processes include: a) discrete slip along strike-slip faults that bound a narrow, highly deformed inner rupture zone; and b), distributed deformation within this inner rupture zone. The latter includes coseismic clockwise rotation of cohesive rafts of turf, soil and near-surface clay-rich sediment. During this process, these “turf rafts” detach from the underlying soil at a mean depth of ~0.7 m, shorten by ~2.5 m (in addition to shortening introduced by any local contractional heave), bulge upwards by < 1 m, and rotate clockwise by ~19° - while also separating from one another along fissures bounded by former (now rotated) synthetic Riedel faults. This rotational deformation accommodated ~3 m of dextral strike-slip (of a total of ~9 m), after which this rotation apparently ceased, regardless of the total slip or the local kinematics (degree of transpression) at any site. The remaining slip was transferred onto later forming, throughgoing faults as discrete displacement. Analysis of the morphology and amplitude of these moletracks suggests that an increase in the degree of transpression (value of contractional heave) at a site increases the magnitude of shortening and the finite longitudinal strain absorbed by the rotated turf rafts, but does not necessarily contribute to an increase in height (generally 0.33-0.53 m on all parts of the fault). Rather, the comparison of these moletracks with those described by other authors suggests that a more controlling factor on their height is the clay content and cohesion of material deformed into the moletracks. Finally, comparison of the before and after cross-sections of the displaced paleoseismic trench has provided, for the first time, insight into how large magnitude strike-slip ruptures are expressed in the fault-orthogonal view typical of paleoseismic trenches. Although this rupture involved ~9 m of dextral strike-slip, the cross-sectional view of the re-excavated trenches was dominated by the much lesser component of fault-perpendicular contractional heave (~1.3 m) that occurred in 2016, which did not occur in previous paleoearthquakes at the same site (these were, by contrast, transtensional). This heave was expressed as up to ~2 m of fault-transverse shortening in the inner rupture zone of the trenches, while the ~9 m of strike-slip only created cm-scale offsets across faults. Previous earthquakes at the site were expressed as cm-dm scale, mostly normal dip-separations of sub-horizontal stratigraphic units across faults, suggesting that a change in local kinematics (of ~8°) must have occurred in 2016. Such a small kinematic change may drastically impact the overall ground expression of strike-slip earthquakes - producing also complicated structures including overprinting fault strands in the rupture zone (to a few metres depth). This information poses challenges for structural geologists and paleoseismologists when interpreting (the significance of) structures in future trench walls.</p>

Accommodating ~9 m of dextral slip on the Kekerengu Fault through ground deformation during the Mw 7.8 Kaikōura Earthquake, November 2016

<p>The Mw 7.8 Kaikōura earthquake of November 14th 2016 provided unprecedented opportunities to understand how the ground deforms during large magnitude strike-slip earthquakes. The re-excavation and extension of both halves of a displaced paleoseismic trench following this earthquake provided an opportunity to test, refine, and extend back in time the known late Holocene chronology of surface rupturing earthquakes on the Kekerengu Fault. As part of this thesis, 28 organic-bearing samples were collected from a suite of new paleoseismic trenches. Six of these samples were added to the preferred age model from Little et al. (2018); this updated age model is now based on 16 total samples. Including the 2016 earthquake, six surface rupturing earthquakes since ~2000 cal. B.P. are now identified and dated on the Kekerengu Fault. Based on the latest five events (E0 to E4), this analysis yields an updated mean recurrence interval estimate for the Kekerengu Fault of 375 ± 32 yrs (1σ) since ~1650 cal. B.P. The older, sixth event (E5) is not included in the preferred model, as it may not have directly preceded E4; however, if this additional event is incorporated into an alternative age model that embraces all six identified events, the mean recurrence interval estimate (considered a maximum) calculated is 433 ± 22 yrs (1σ) since ~2000 cal. B.P. Comparison of structures on an identical trench wall logged both before and after the 2016 earthquake, and analysis of pre- and post-earthquake high resolution imagery and Digital Surface Models (DSMs), has allowed the quantification of where and how ~9 m of dextral-oblique slip was accommodated at this site during the earthquake. In addition to this, I analyse the coseismic structure of the adjoining segment of the 2016 ground rupture using detailed post-earthquake aerial orthophotography, to further investigate how geological surface structures (bulged-up moletrack structures) accommodated slip in the rupture zone. These combined analyses allowed me to identify two primary deformation mechanisms that accommodated the large coseismic slip of this earthquake, and the incremental effect of that slip on the structural geology of the rupture zone. These processes include: a) discrete slip along strike-slip faults that bound a narrow, highly deformed inner rupture zone; and b), distributed deformation within this inner rupture zone. The latter includes coseismic clockwise rotation of cohesive rafts of turf, soil and near-surface clay-rich sediment. During this process, these “turf rafts” detach from the underlying soil at a mean depth of ~0.7 m, shorten by ~2.5 m (in addition to shortening introduced by any local contractional heave), bulge upwards by < 1 m, and rotate clockwise by ~19° - while also separating from one another along fissures bounded by former (now rotated) synthetic Riedel faults. This rotational deformation accommodated ~3 m of dextral strike-slip (of a total of ~9 m), after which this rotation apparently ceased, regardless of the total slip or the local kinematics (degree of transpression) at any site. The remaining slip was transferred onto later forming, throughgoing faults as discrete displacement. Analysis of the morphology and amplitude of these moletracks suggests that an increase in the degree of transpression (value of contractional heave) at a site increases the magnitude of shortening and the finite longitudinal strain absorbed by the rotated turf rafts, but does not necessarily contribute to an increase in height (generally 0.33-0.53 m on all parts of the fault). Rather, the comparison of these moletracks with those described by other authors suggests that a more controlling factor on their height is the clay content and cohesion of material deformed into the moletracks. Finally, comparison of the before and after cross-sections of the displaced paleoseismic trench has provided, for the first time, insight into how large magnitude strike-slip ruptures are expressed in the fault-orthogonal view typical of paleoseismic trenches. Although this rupture involved ~9 m of dextral strike-slip, the cross-sectional view of the re-excavated trenches was dominated by the much lesser component of fault-perpendicular contractional heave (~1.3 m) that occurred in 2016, which did not occur in previous paleoearthquakes at the same site (these were, by contrast, transtensional). This heave was expressed as up to ~2 m of fault-transverse shortening in the inner rupture zone of the trenches, while the ~9 m of strike-slip only created cm-scale offsets across faults. Previous earthquakes at the site were expressed as cm-dm scale, mostly normal dip-separations of sub-horizontal stratigraphic units across faults, suggesting that a change in local kinematics (of ~8°) must have occurred in 2016. Such a small kinematic change may drastically impact the overall ground expression of strike-slip earthquakes - producing also complicated structures including overprinting fault strands in the rupture zone (to a few metres depth). This information poses challenges for structural geologists and paleoseismologists when interpreting (the significance of) structures in future trench walls.</p>

Deep Coseismic Slip in the Cascadia Megathrust can be Consistent with Coastal Subsidence

At subduction zones, the down-dip limit of slip represents how deep an earthquake can rupture. For hazards it is important - it controls the intensity of shaking and the pattern of coseismic uplift and subsidence. In the Cascadia Subduction Zone, because no large magnitude events have been observed in instrumental times, the limit is inferred from geological estimates of coastal subsidence during previous earthquakes; it is typically assumed to coincide approximately with the coastline. This is at odds with geodetic coupling models, it leaves residual slip deficits unaccommodated on a large swath of the megathrust. Here we will show that ruptures can penetrate deeper into the megathrust and still produce coastal subsidence provided slip decreases with depth. We will discuss the impacts of this on expected shaking intensities

Effect of the Kaikōura Earthquake on Velocity Changes In and Around the Ruptured Region: A Noise Cross-Correlation Approach

<p>Seismic velocity changes before and after large magnitude earthquakes carry information about damage present within the faults in the surrounding region. In this thesis, temporal velocity changes are measured before and after the 2016 Kaikōura earthquake using ambient noise interferometry between 2012 - 2018. This period contains the Mw 7.8 2016 Kaikoura earthquake as well as the 2013 Cook Strait earthquake sequence and a few deep large magnitude earthquakes in 2015 - 2016. Three primary objectives are identified: (1) investigate seismic velocity changes in the Kaikōura region and their connection to the 2016 Kaikōura earthquake to try and determine if there was a change before/after the earthquake, (2) determine how this change varied across the region, and (3) consider if ambient noise can lead to improved detection and understanding of geological hazard. The primary approach used to measure velocity changes in the Kaikōura region involved cross correlating noise recorded by seismic stations across the region. Velocity changes are sought by averaging the best result from multiple onshore station pairs. A secondary approach was also used, in which specific station pairs were averaged to determine if there were more localised velocity changes over more specific regions. This was to determine if the velocity changes observed following the 2016 Kaikōura earthquake occurred over the entire ruptured region. Following the 2016 Kaikōura earthquake a velocity decrease of 0.24±0.02% was observed on the average of the vertical-vertical components for eight stations. The remaining eight cross-component pairs showed a smaller seismic decrease with an average value of 0.22±0.05%. After the decrease following the Kaikōura earthquake, there is a steady velocity increase of 0.13±0.02% over a one-and-a-half-year period. This indicates that prior to the earthquake, seismic velocity was at a steady state until it was perturbed by the Kaikōura earthquake, and seismic velocities rapidly decreased over all stations. Across the region, stations with a longer interstation distance and further away from ruptured faults had a smaller decrease in velocity than station pairs with a smaller interstation distance that were closer to ruptured faults. We interpret the velocity decrease following the Kaikōura earthquake as a result of cracks opening during the earthquake. The velocity increase following the earthquake is indicative of the cracks slowly healing. The Cook Strait earthquake sequence that occurred in 2013 did not cause any velocity changes at the stations used in this thesis. This has been interpreted to be because the changes were too small compared to the background noise or the stations were not recording during the time of the earthquake sequence. Two other decreases were also observed in the region following two deep earthquakes in April 2015 (Mw 6.2, depth = 52km) and February 2016 (Mw 5.7, depth = 48km). Both of these events resulted in a small seismic decrease of 0.1±0.02%. Although these earthquakes were close to seismic stations when they occurred, they were much deeper and had a smaller magnitude than the Kaikōura earthquake so did not cause a large velocity decrease. By understanding what causes velocity changes it is possible to have an improved understanding of the geological hazard in the region.</p>

Effect of the Kaikōura Earthquake on Velocity Changes In and Around the Ruptured Region: A Noise Cross-Correlation Approach

<p>Seismic velocity changes before and after large magnitude earthquakes carry information about damage present within the faults in the surrounding region. In this thesis, temporal velocity changes are measured before and after the 2016 Kaikōura earthquake using ambient noise interferometry between 2012 - 2018. This period contains the Mw 7.8 2016 Kaikoura earthquake as well as the 2013 Cook Strait earthquake sequence and a few deep large magnitude earthquakes in 2015 - 2016. Three primary objectives are identified: (1) investigate seismic velocity changes in the Kaikōura region and their connection to the 2016 Kaikōura earthquake to try and determine if there was a change before/after the earthquake, (2) determine how this change varied across the region, and (3) consider if ambient noise can lead to improved detection and understanding of geological hazard. The primary approach used to measure velocity changes in the Kaikōura region involved cross correlating noise recorded by seismic stations across the region. Velocity changes are sought by averaging the best result from multiple onshore station pairs. A secondary approach was also used, in which specific station pairs were averaged to determine if there were more localised velocity changes over more specific regions. This was to determine if the velocity changes observed following the 2016 Kaikōura earthquake occurred over the entire ruptured region. Following the 2016 Kaikōura earthquake a velocity decrease of 0.24±0.02% was observed on the average of the vertical-vertical components for eight stations. The remaining eight cross-component pairs showed a smaller seismic decrease with an average value of 0.22±0.05%. After the decrease following the Kaikōura earthquake, there is a steady velocity increase of 0.13±0.02% over a one-and-a-half-year period. This indicates that prior to the earthquake, seismic velocity was at a steady state until it was perturbed by the Kaikōura earthquake, and seismic velocities rapidly decreased over all stations. Across the region, stations with a longer interstation distance and further away from ruptured faults had a smaller decrease in velocity than station pairs with a smaller interstation distance that were closer to ruptured faults. We interpret the velocity decrease following the Kaikōura earthquake as a result of cracks opening during the earthquake. The velocity increase following the earthquake is indicative of the cracks slowly healing. The Cook Strait earthquake sequence that occurred in 2013 did not cause any velocity changes at the stations used in this thesis. This has been interpreted to be because the changes were too small compared to the background noise or the stations were not recording during the time of the earthquake sequence. Two other decreases were also observed in the region following two deep earthquakes in April 2015 (Mw 6.2, depth = 52km) and February 2016 (Mw 5.7, depth = 48km). Both of these events resulted in a small seismic decrease of 0.1±0.02%. Although these earthquakes were close to seismic stations when they occurred, they were much deeper and had a smaller magnitude than the Kaikōura earthquake so did not cause a large velocity decrease. By understanding what causes velocity changes it is possible to have an improved understanding of the geological hazard in the region.</p>