Network Geometry Research Articles

Distributed cohesive configuration controller for a swarm with low‐cost platforms

AbstractThis study presents a cohesive configuration controller for distributed space coverage by a swarm of robots. The goal is to build a dense, convex network that is robust against disconnection while robots are flocking with only incomplete knowledge about the network. The controller is an integrated framework of two different algorithms. First, we present a boundary force algorithm: physics‐based swarm intelligence that borrows the concept of surface tension force between liquid molecules. The combination of such a force with conventional flocking produces a convex and dense configuration without knowledge of the complete geometry of a robot network. Second, robots distributively determine when a configuration is on the verge of disconnection by identifying a local articulation point—a region where the removal of a single robot will change the local topology. When such a point is detected, robots switch their behavior to clustering, which aggregates them around the vulnerable region to remove every articulation point and retain a connected configuration. Finally, we introduced an index that objectively represents the level of risk of a robot configuration against the massive fragmentation, called vulnerability index. We provide theoretical performance analyses of each algorithm and validate the results with simulations and experiments using a set of low‐cost robots.

A Machine-Learning Approach for the Reconstruction of Ground-Shaking Fields in Real Time

ABSTRACT Real-time seismic monitoring is of primary importance for rapid and targeted emergency operations after potentially destructive earthquakes. A key aspect in determining the impact of an earthquake is the reconstruction of the ground-shaking field, usually expressed as the ground-motion parameter. Traditional algorithms compute the ground-shaking field from the punctual data at the stations relying on ground-motion prediction equations computed on estimates of the earthquake location and magnitude when the instrumental data are missing. The results of such algorithms are then subordinate to the evaluation of location and magnitude, which can take several minutes. To fill the temporal gap between the arrival of the data and the estimate of these parameters, a new data-driven algorithm that exploits the information from the station data only is introduced. This algorithm, consisting of an ensemble of convolutional neural networks (CNNs) trained on a database of ground-shaking maps produced with traditional algorithms, can provide estimates of the ground-shaking maps and their associated uncertainties in real time. Because CNNs cannot handle sparse data, a Voronoi tessellation of a selected peak ground parameter recorded at the stations is computed and used as the input to the CNNs; site effects and network geometry are accounted for using a (normalized) VS30 map and a station location map, respectively. The developed method is robust to noise, can handle network geometry changes over time without the need for retraining, and can resolve multiple simultaneous events. Although having a lower resolution, the results obtained are statistically compatible with the ones from traditional methods. A fully operational version of the algorithm is running on the servers at the Department of Mathematics and Geosciences of the University of Trieste, showing real-time capabilities in handling stations from multiple Italian strong-motion networks and outputting results with a resolution of 0.05° × 0.05°.

Identifying the skeptics and the undecided through visual cluster analysis of local network geometry

By skeptics and undecided we refer to nodes in clustered social networks that cannot be assigned easily to any of the clusters. Such nodes are typically found either at the interface between clusters (the undecided) or at their boundaries (the skeptics). Identifying these nodes is relevant in marketing applications like voter targeting, because the persons represented by such nodes are often more likely to be affected in marketing campaigns than nodes deeply within clusters. So far this identification task is not as well studied as other network analysis tasks like clustering, identifying central nodes, and detecting motifs. We approach this task by deriving novel geometric features from the network structure that naturally lend themselves to an interactive visual approach for identifying interface and boundary nodes.

Tropical approximation to finish time of activity networks.

We breakdown complex projects into activities and their logical dependencies. We estimate the project finish time based on the activity durations and relations. However, adverse events trigger delay cascades shifting the finish time. Here I derive a tropical algebraic equationfor the finish time of activity networks, encapsulating the principle of linear superposition of exogenous perturbations in the tropical sense. From the tropical algebraic equationI derive the finish time distribution with explicit reference to the distribution of exogenous delays and the network topology and geometry.

Tracer-aided identification of hydrological and biogeochemical controls on in-stream water quality in a riparian wetland

In-stream water quality reflects the integrated results of hydrological mixing of different water sources and associated biogeochemical transformations. However, quantifying the relative importance of these controls is often challenging, particularly in riparian wetlands due to complex process interactions and marked spatio-temporal heterogeneity in environmental gradients. Here, we established a two-step method to differentiate the dominance of hydrological and biogeochemical controls on water quality in a riparian peatland in northern Germany. First, an isotope-based mixing model was developed for distributed modelling of in-stream water balance over a two-year period. The simulation showed the predominance of groundwater inflows for most of the time period, while lateral inflows and channel leakage became more influential in mid-summer, as stream-groundwater connectivity weakened due to declining groundwater levels. A moderate downstream shift from groundwater to lateral inflow was also observed due to the changing channel network geometries and inflow from field drains. The mixing model was then further applied to predict the in-stream concentrations of nutrients, major ions and trace elements. The predicted concentrations were assumed to be those resulting from hydrological mixing only, while influence of biogeochemical controls were reflected by the prediction deviation from observation. Accordingly, 15 water quality parameters were grouped based on their simulation performances into hydrologically-controlled (Cl–, Mg, Na, K, and Si), biogeochemically-controlled (DOC, SO42−, Mn, and Zn), or controlled-by-both (SRP, NO3-N, Ca, Fe, Al, and Cu). The mixing modelling not only reproduced the spatiotemporal in-stream water balance with finer process conceptualisation, but also provided a generic method to quantitatively disentangle the relative strength of hydrological and biogeochemical controls. Such a method can be employed as a robust learning tool before extending a hydrological model for water quality simulation, as when, where and how strong biogeochemical controls are exerted provides a strong indicator on which dominant processes need to be conceptualised.

Uncertainty propagation analysis of the computed ITER torus effective pumping speed during the dwell phase

At the University of Thessaly the ARIADNE code for modeling complex gas distribution systems operating under any vacuum conditions has been developed by integrating a kinetic database to a typical gas network solver. The ARIADNE code has been successfully implemented to model the ITER primary pumping system providing the torus effective pumping speed, as well as the pressure evolution during the dwell phase. However, the computed results are subject to the input data, which include the pipe network geometry, approximating the real geometry of the ITER primary pumping system and the operating data, such as the torus pressure, gas temperature and cryopump pumping speed. The effect of the aforementioned input quantity uncertainties to the torus effective pumping speed is investigated via an uncertainty propagation analysis by coupling the Monte Carlo method (MCM) with the ARIADNE code. Documenting the propagation of each input parameter uncertainty to the torus effective pumping speed uncertainty is beneficial for judging the accuracy of the modeling and simulation results, as well as for identifying the most important sources of uncertainty. Furthermore, the presented methodology can be used to investigate the uncertainty propagation of any input quantity to any output quantity for vacuum systems of arbitrary complexity.

Structural Features of Microvascular Networks Trigger Blood Flow Oscillations

We analyse mathematical models in order to understand how microstructural features of vascular networks may affect blood flow dynamics, and to identify particular characteristics that promote the onset of self-sustained oscillations. By focusing on a simple three-node motif, we predict that network “redundancy”, in the form of a redundant vessel connecting two main flow-branches, together with differences in haemodynamic resistance in the branches, can promote the emergence of oscillatory dynamics. We use existing mathematical descriptions for blood rheology and haematocrit splitting at vessel branch-points to construct our flow model; we combine numerical simulations and stability analysis to study the dynamics of the three-node network and its relation to the system’s multiple steady-state solutions. While, for the case of equal inlet-pressure conditions, a “trivial” equilibrium solution with no flow in the redundant vessel always exists, we find that it is not stable when other, stable, steady-state attractors exist. In turn, these “nontrivial” steady-state solutions may undergo a Hopf bifurcation into an oscillatory state. We use the branch diameter ratio, together with the inlet haematocrit rate, to construct a two-parameter stability diagram that delineates regimes in which such oscillatory dynamics exist. We show that flow oscillations in this network geometry are only possible when the branch diameters are sufficiently different to allow for a sufficiently large flow in the redundant vessel, which acts as the driving force of the oscillations. These microstructural properties, which were found to promote oscillatory dynamics, could be used to explore sources of flow instability in biological microvascular networks.

DevelNet: Earthquake Detection on Develocorder Films with Deep Learning: Application to the Rangely Earthquake Control Experiment

Abstract There exists over a century of instrumental seismic data; however, most seismograms recorded before the 1980s are only available in analog form. Although analog seismograms are of great value, they are underutilized due to the difficulties of making quantitative measurements on the original media and in converting them to digital time series. In this study, we present an alternative workflow, based on deep learning, to reconstruct an earthquake catalog from images of analog data without conversion to vector time series. We trained a convolutional neural network—DevelNet, using synthetic analog data to detect earthquakes on scanned multichannel Develocorder film images. We then developed an image-based processing workflow to measure arrival times, locate, and determine the magnitudes of earthquakes in the data. We demonstrate the performance of this approach on two years of continuous Develocorder film recordings from the Rangely earthquake control experiment in the mid-1970s. Our approach detects twice the number of events reported in the original catalog (Raleigh et al., 1976). This demonstrates that DevelNet efficiently detects earthquakes from Develocorder film scans, performs consistently over time, and is robust to changes in network geometry. Our locations generally agree with the original study, although the automatically measured arrival times are less precise than manual reading, leading to increased location scatter. Our automatic workflow of Develocorder films rivals the performance of skilled analysts in earthquake detection, but with minimal human intervention. This image-based processing offers a new approach for effectively and efficiently extracting earthquake information from analog seismic data.

Pore network and solute flux pattern analysis towards improved predictability of diffusive transport in argillaceous host rocks

Clay rock formations are considered as host rocks for underground radioactive waste repositories. Reliable predictions of diffusive transport heterogeneity are critical for assessing the sealing capacity of argillaceous rocks. The predictive power of numerical approaches to flow field analysis and radionuclide migration depends on the quality of the underlying pore network geometry. Both sedimentary and diagenetic complexity are controlling factors.In this study, we demonstrate a cross-scale approach to reconstruct the pore network geometries of the sandy facies of the Opalinus Clay rock. We identified diagenetic and sedimentary subfacies components based on the concentration of diagenetic carbonates and sulfides and grain size variability, and quantified their pore size distributions and pore network geometries. A viable approach for use in transport modeling is to combine μ-CT data segmentation followed by filling the resulting volumes with representative pore network geometries based on FIB-SEM data. The resulting generalized pore network geometries are applied in digital rock models to calculate effective diffusivities, using a combined upscaling workflow for transport simulations from nanometer to micrometer scales.Positron emission tomography (PET) diffusion experiments validated the transport simulation results. We introduced a statistical treatment of the PET and μ-CT tomographic datasets based on the spatial variability of both PET tracer concentrations and rock density. The analyzed effective diffusivities confirmed the numerical results.This study illustrates three important steps in migration analysis: (i) a workflow of general applicability for cross-scale identification of pore network data in argillaceous rocks, (ii) application of the pore network data for the numerical analysis of diffusive transport, and (iii) validation of numerical results via combined PET - μ-CT diffusion experiments. Although the conceptual approach is not feasible for large numbers of samples, it opens up a strong potential for generalization: the validated results of effective diffusivities can now be easily used in a variety of segmented geometries. This allows to efficiently test upscaling concepts for the continuum scale on this basis.

The role of fracture branching in the evolution of fracture networks: An outcrop study of the Jurassic Navajo Sandstone, southern Utah

Fractures strongly influence the permeability of geologic formations, and because most fractures in the subsurface are below the resolution of geophysical methods, predicting the spatial evolution of fracture networks is important for groundwater resources, oil and gas production, and geothermal energy. Previous researchers have established that variations in lithologic mechanical properties influence the propagation of joints and fractures in layered rocks under stable stress conditions, but few studies have addressed how instabilities in local stress fields, in conjunction with variations in rock mechanical properties, lead to fracture branching.We investigate NE-striking fractures in the footwall of the west-dipping Sevier fault in Utah, USA, where canyons expose the sub-horizontal Jurassic Navajo Sandstone. We analyzed field data, petrographic data, and unmanned-aerial-vehicle (UAV) imagery to document the fracture network. We also used structure-from-motion (SfM) software to build a 3D virtual outcrop model, measuring 100 m high and 260 m wide, to aid our analysis of fracture network geometry, intensity, and spacing variations.Data reveal an up-section increase in fracture intensity and decrease in spacing regularity partly accommodated by upward fracture branching. We suggest that branching was initiated by a combination of changes in mechanical behavior within cross-bed sets and twist hackle propagation associated with mixed mode fracture systems. These tree-like fracture geometries may be associated with high-velocity, earthquake-related fracture propagation events. Fracture branching strongly influences permeability and should be considered by researchers investigating fracture development in subsurface systems.

The impact of changes in natural fracture fluid pressure on the creation of fracture networks

Complex fracture networks have been observed in mine-back experiments, core samples, and other fracture diagnostic measurements. The interaction between hydraulic and natural fractures plays a key role in the formation of this complexity in naturally fractured formations. In past studies, changes in fluid pressure in natural fractures induced by interaction with other fractures are neglected when interactions between natural and hydraulic fractures are considered. In this paper, a DDM-based hydraulic fracturing simulator is developed to investigate the effect of changes in fluid pressure in natural fractures on fracture behavior and the resulting geometry of complex fracture networks. A fully implicit method is used to solve for width and pressure simultaneously. The proposed model is validated against an analytical solution for such variations in the vicinity of a penny-shaped fracture. Two kinds of interactions between hydraulic and natural fractures are defined to clearly investigate the mechanism of the far-field failure of natural fractures. Comparisons with and without considering changes in fluid pressure in natural fractures are conducted to show how the geometry of hydraulic fractures changes when changes in fluid pressure in fractures are included. Key parameters such as in-situ stress, natural fracture orientation, and friction coefficient are varied to develop new failure and crossing criteria under different conditions.Simulation results show that (1) changes in fluid pressure in fractures can initiate isolated fractures to create a stimulated rock volume. In some instances this will result in a more branched and complex fracture network; (2) deflection of hydraulic fractures along natural fractures is promoted when the effect of changes in fluid pressure in natural fractures is taken into account; (3) hydraulic fractures tend to deflect along pre-existing natural fractures with no far-field failure when the stress contrast is low; (4) in formations with high in-situ stress contrast, far-field failure of natural fractures occurs more readily and is impacted by the orientation and frictional coefficient of the natural fractures; (5) as expected, well cemented (highly mineralized) fractures with larger frictional coefficients hinder the slip of natural fractures and result in more planar less branched fractures. The results for failure and crossing criteria we developed can improve the understanding of the formation of complex fracture networks in naturally fractured rocks.

Branched actin networks are organized for asymmetric force production during clathrin-mediated endocytosis in mammalian cells

Actin assembly facilitates vesicle formation in several trafficking pathways, including clathrin-mediated endocytosis (CME). Interestingly, actin does not assemble at all CME sites in mammalian cells. How actin networks are organized with respect to mammalian CME sites and how assembly forces are harnessed, are not fully understood. Here, branched actin network geometry at CME sites was analyzed using three different advanced imaging approaches. When endocytic dynamics of unperturbed CME sites are compared, sites with actin assembly show a distinct signature, a delay between completion of coat expansion and vesicle scission, indicating that actin assembly occurs preferentially at stalled CME sites. In addition, N-WASP and the Arp2/3 complex are recruited to one side of CME sites, where they are positioned to stimulate asymmetric actin assembly and force production. We propose that actin assembles preferentially at stalled CME sites where it pulls vesicles into the cell asymmetrically, much as a bottle opener pulls off a bottle cap.

Morphological controls on surface runoff: an interpretation of steady-state energy patterns, maximum power states and dissipation regimes within a thermodynamic framework

Abstract. Recent research explored an alternative energy-centred perspective on hydrological processes, extending beyond the classical analysis of the catchment's water balance. Particularly, streamflow and the structure of river networks have been analysed in an energy-centred framework, which allows for the incorporation of two additional physical laws: (1) energy is conserved and (2) entropy of an isolated system cannot decrease (first and second law of thermodynamics). This is helpful for understanding the self-organized geometry of river networks and open-catchment systems in general. Here we expand this perspective, by exploring how hillslope topography and the presence of rill networks control the free-energy balance of surface runoff at the hillslope scale. Special emphasis is on the transitions between laminar-, mixed- and turbulent-flow conditions of surface runoff, as they are associated with kinetic energy dissipation as well as with energy transfer to eroded sediments. Starting with a general thermodynamic framework, in a first step we analyse how typical topographic shapes of hillslopes, representing different morphological stages, control the spatial patterns of potential and kinetic energy of surface runoff and energy dissipation along the flow path during steady states. Interestingly, we find that a distinct maximum in potential energy of surface runoff emerges along the flow path, which separates upslope areas of downslope potential energy growth from downslope areas where potential energy declines. A comparison with associated erosion processes indicates that the location of this maximum depends on the relative influence of diffusive and advective flow and erosion processes. In a next step, we use this framework to analyse the energy balance of surface runoff observed during hillslope-scale rainfall simulation experiments, which provide separate measurements of flow velocities for rill and for sheet flow. To this end, we calibrate the physically based hydrological model Catflow, which distributes total surface runoff between a rill and a sheet flow domain, to these experiments and analyse the spatial patterns of potential energy, kinetic energy and dissipation. This reveals again the existence of a maximum of potential energy in surface runoff as well as a connection to the relative contribution of advective and diffusive processes. In the case of a strong rill flow component, the potential energy maximum is located close to the transition zone, where turbulence or at least mixed flow may emerge. Furthermore, the simulations indicate an almost equal partitioning of kinetic energy into the sheet and the rill flow component. When drawing the analogy to an electric circuit, this distribution of power and erosive forces to erode and transport sediment corresponds to a maximum power configuration.

Blending fiber-shaped long-range conductive additives for better battery performance: Mechanism study based on heterogeneous electrode model

Conductive additives in lithium-ion batteries are indispensable for constructing electronic conductive networks and enabling continuous electrochemical reactions. Fine carbon black powder, typically composed of small round particles with a diameter of tens of nanometers, is the most commonly used conductive additive. The experimental results in some literature indicate that blending fiber-shaped conductive additives can help boost battery performance. The heterogenous electrode model is employed in this study to explore the distribution of electronic current density within lithium-ions batteries, demonstrate the benefits of blending fiber-shaped long-range conductive additives, and shed light on its mechanism. Our findings show that the introduction of long-range conductive additives leads to reallocation of the electron flow: as the proportion/weight ratio of long-range conductive additives increases, the electronic current undertaken by all conductive additives increases, while that undertaken by the active particles decreases. Without changing the total wt.% of conductive additives in batteries, the ohmic resistance of the electrode and the whole cell can be significantly reduced due to more electrons passing through compositions with better conductivity. These results help researchers gain a better understanding on the correlation between conductive network geometry and electron transfer effectiveness.

Dissociation of polymeric micelle under hemodynamic shearing

Many studies have observed the unexpected micelle dissociation occurred upon i.v. injection. However the dynamics and mechanism of in vivo micelle dissociation are still unclear, mainly due to the lack of physiologically representative models. Here, we used microfluidic channels to mimic geometries of vascular networks and related hemodynamic shearing conditions, adopted fluorescence resonance energy transfer (FRET) imaging to monitor the dynamics of the micelle dissociation and applied the dissipative particle dynamics (DPD) to simulate the morphological evolution of micelles under shearing. In vessel-mimicking microfluidic models, we observed the fast dissociation of clinically relevant polyethylene glycol-block-poly(ε-caprolactone) (PEG-PCL) and PEG-block-poly(D,L-lactide) (PEG-PDLLA) micelles that were stable under static conditions. FRET imaging from a pair of fluorophores (Cy5 and Cy5.5) conjugated in the micelle core revealed that the dynamics of micelle dissociation was associated with hemodynamic shearing, which was altered by either tuning the flow rate of mouse blood or changing the geometry of the microchannel. In addition to blood proteins that were generally considered as the major contributors to the micelle dissociation, we found in blood flow, the presence of shear field on particles, as surrogates to red blood cells, significantly influenced the micelle dissociation. Moreover, the DPD stimulation revealed that the morphological evolution of micelles under shearing resulted in the disintegrity of the protective PEG shell, leading to the increased exposure of the hydrophobic core to the outer media, dramatically facilitating the nearby blood components to interact with the inner core and therefore quickly destructed the micellar structures. These findings suggested the mechanism of the shear-induced micelle dissociation in blood flow, which can be directional for the design of micellar nanomedicine with expected circulation lifespan, and therefore high therapeutic efficacy.

Coarse Graining on Financial Correlation Networks

Community structure detection is an important and valuable task in financial network studies as it forms the basis of many statistical applications such as prediction, risk analysis, and recommendation. Financial networks have a natural multi-grained structure that leads to different community structures at different levels. However, few studies pay attention to these multi-part features of financial networks. In this study, we present a geometric coarse graining method based on Voronoi regions of a financial network. Rather than studying the dense structure of the network, we perform our analysis on the triangular maximally filtering of a financial network. Such filtered topology emerges as an efficient approach because it keeps local clustering coefficients steady and it underlies the network geometry. Moreover, in order to capture changes in coarse grains geometry throughout a financial stress, we study Haantjes curvatures of paths that are the farthest from the center in each of the Voronoi regions. We performed our analysis on a network representation comprising the stock market indices BIST (Borsa Istanbul), FTSE100 (London Stock Exchange), and Nasdaq-100 Index (NASDAQ), across three financial crisis periods. Our results indicate that there are remarkable changes in the geometry of coarse grains.

Local Ice-like Structure at the Liquid Water Surface.

Experiments and computer simulations have established that liquid water's surfaces can deviate in important ways from familiar bulk behavior. Even in the simplest case of an air-water interface, distinctive layering, orientational biases, and hydrogen bond arrangements have been reported, but an overarching picture of their origins and relationships has been incomplete. Here we show that a broad set of such observations can be understood through an analogy with the basal face of crystalline ice. Using simulations, we demonstrate a number of structural similarities between water and ice surfaces, suggesting the presence of domains at the air-water interface with ice-like features that persist over 2-3 molecular diameters. Most prominent is a shared characteristic layering of molecular density and orientation perpendicular to the interface. Lateral correlations of hydrogen bond network geometry point to structural similarities in the parallel direction as well. Our results bolster and significantly extend previous conceptions of ice-like structure at the liquid's boundary and suggest that the much-discussed quasi-liquid layer on ice evolves subtly above the melting point into a quasi-ice layer at the surface of liquid water.

Numerical investigation of the fracture network morphology in multi-cluster hydraulic fracturing of horizontal wells: A DDM-FVM study

Hydraulic fracturing is the most effective method to enhance the recovery of unconventional oil and gas via the formation of complex connected fracture networks. In this study, a pseudo three-dimensional fully fluid-solid coupled model is established to investigate the interaction between hydraulic fractures (HFs) and natural fractures (NFs), and the fracture network propagation process. The displacement discontinuity method (DDM) and finite volume method (FVM) are applied to simulate the rock deformation of the reservoir and fracturing fluid flow into the fracture networks, respectively. The influence of the stress shadow, flow rate distribution, comprehensive fluid loss, and complex extension behavior are considered in this model. A modified analytical crossing criterion is used to predict the interaction between the HF and NF. To improve the accuracy of the pressure calculation, the width of the fracture vertical cross-section is replaced by an equivalent width. The coupled equations are solved by Newton–Raphson iteration method. The numerical results of the proposed model agreed well with published analytical and experimental results. Based on the model, detailed sensitivity analyses are conducted to investigate the interaction between HFs and NFs, and the geometries of the fracture networks. The simulation results show that the fracture network morphology is significantly affected by NFs in naturally fractured reservoirs. The stress interference between HFs can weaken the effect of the horizontal stress difference; thus, NFs can be activated effectively. In addition, the complexity of the fracture network is proportional to the activation rate of the NFs. A more complex fracture network can be formed under the following conditions: a small horizontal stress difference, large NF length, low injection rate and low viscosity. It is noteworthy that a smaller value is not always better for cluster spacing. An optimal cluster spacing exist to obtain good results for reservoir stimulation.

Self-organized nanoscale networks: are neuromorphic properties conserved in realistic device geometries?

Abstract Self-organised nanoscale networks are currently under investigation because of their potential to be used as novel neuromorphic computing systems. In these systems, electrical input and output signals will necessarily couple to the recurrent electrical signals within the network that provide brain-like functionality. This raises important questions as to whether practical electrode configurations and network geometries might influence the brain-like dynamics. We use the concept of criticality (which is itself a key charactistic of brain-like processing) to quantify the neuromorphic potential of the devices, and find that in most cases criticality, and therefore optimal information processing capability, is maintained. In particular we find that devices with multiple electrodes remain critical despite the concentration of current near the electrodes. We find that broad network activity is maintained because current still flows through the entire network. We also develop a formalism to allow a detailed analysis of the number of dominant paths through the network. For rectangular systems we show that the number of pathways decreases as the system size increases, which consequently causes a reduction in network activity.

The role of turns in pedestrian route choice: A clarification

Among a number of variables shown to affect pedestrian route choice, path length and turns have stood out as the most consequential. Turns have been considered the superior variable by some architectural scholars of urban street networks, while transportation planners and geographers believe distance to be paramount. The longstanding debate between these two approaches has been reinvigorated with the emergence of big data and advanced computational methods. In this paper, we provide much-needed clarity to this debate by demonstrating how the relative effect of turns depends on the spatial properties of street networks. We postulate that certain properties of street networks make it possible to reduce the number of turns without substantially increasing route distance, and suggest that data from such environments are likely to show a large effect of turns on route choice. Conversely, networks where a reduction of in turns is necessarily accompanied by an increase in distance are expected to reduce the effect of turns. We test these hypotheses by examining the effects of both distance and turns on pedestrian route choice using a path size logit model calibrated on over 10,000 anonymized GPS traces of pedestrians each in San Francisco, CA and Boston, MA. We find that the effect of distance is consistently larger in magnitude, while the effect of turns depends on network geometry. Only in specific street networks can turns alone explain route choice behavior as well as distance. Our findings suggest that turns, as well as other environmental qualities of a route, should be considered in addition to, not in lieu of, distance.