Kung Fu Patterns

[Jenifer Griffard]

Withsummer finally here, the redfish activity in the grass flats should start to pick up. With this in mind, I thought I would share with you one of my favorite saltwater patterns for flats feeding redfish. This pattern I believe is from Eddie Wyatt. Since this patterns inception, it has been tied in several variations and I believe this may be another. This pattern is a fairly easy fly to tie and very durable. It can be tied in many different color variations to imitate the crabs in the flats you are fishing. I like this fly tied in traditional blue crab coloration. I also like to tie it with a heavy hourglass eye so that it will drop through the grass.

The weed guard is optional although, I tie almost all of my crab patterns for flats fish weed less now. The most difficult part of this fly is probably working with the EP Fibers. To manage things, a very small hair clamp works great. Lastly, the eyes can either be purchased or made fairly easily with hard mono melted, painted with fingernail polish and coated with CCG Tack Free or your choice of tack free uv resin. As always, best wishes and tight lines!

Copyright: 2014 Kung et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: Audi, Volkswagen, BBVA, The Coca Cola Company, Ericsson, Ferrovial, GE provided funding for this study. There are no patents, products in development or marketed products to declare. This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Of specific interest in this domain is the study of human mobility in the context of our commute behaviors, as insights from such pursuits often have potent and far-reaching implications in urban planning, infrastructure construction, and even epidemiology. The first forays into this area using CDRs come from Becker et al. [12], who used the bulk mobile network data to understand the daily and nightly profiles of activities in Morristown, NJ. More recently, Issacman et al. [13] undertook a comparative study of daily commute patterns over two U.S. cities (New York and Los Angeles). While these studies have laid the groundwork for some key insights into the behavioral patterns in human commuting, they have been rather limited in scope. By focusing on a few select cities specifically in the U.S., the similarities and differences observed are perhaps more accountable to regional determinants, as opposed to fundamental cultural and/or evolutionary factors. If we really wish to understand the characteristics of human commute patterns, we also need to set our eyes more broadly to countrywide datasets that come from different parts of the world. As a proof of concept, we have focused on Portugal (in Europe) and Ivory Coast (in Sub-Saharan Africa) for our study.

In the Milan GPS dataset, the home/work locations are estimated in a similar manner, with the assumption that the individual always stays in proximity to his/her car. This assumption is not always valid, as the individual may park his/her car and run several errands at the same time. Despite potential inaccuracies that arise due to these behaviors, the nature of the GPS dataset makes accounting for such behaviors impossible, unless the same dataset can be overlaid user-by-user against other mobility-related datasets (such as a CDR dataset of Milan, which was not available to us).

Despite these filtering and correction measures, there are still some potential limitations about the datasets that can negatively affect the accuracy of our results and their interpretation. Such included the differences in the mobile phone usage behaviors across different users, and notably, across different countries/cultures. People who call at different times of the day with different frequencies, for example, can affect our estimations of the commute times in different locations. There exist more elaborate correction mechanisms that we may employ to further screen for these confounding factors. However, given the limited size of the dataset already (for example, for Ivory Coast, about 500,000 total users, each traced over only 2 weeks), such more elaborate methodologies are beyond the scope of our study. However, in the Discussion section, we review in greater detail the potential impacts that these limitations may have on the accuracy of our study of human commute behaviors.

Having developed a common way to parse for home/work and commute information that can be equally applied to different datasets, we can then ask what insights this methodology can reveal to us regarding human mobility and commuting. In this section, we discuss a few topics/insights about the datasets that stem from our methodology, and conclude by focusing on testing the constant travel time budget hypothesis in the context of commuting.

There are some concerns of whether or not the distribution of cell towers in our datasets offer sufficient spatial resolution to interrogate human commuting behaviors, especially for countrywide datasets such as Ivory Coast, Portugal, and Saudi Arabia. Figure 1 in a previous study by Amini et al. [42] attempted to characterize this spatial distribution in detail. Figure S2(c) in this paper plots the cell tower density for Saudi Arabia. Notably, the cell tower spacing is not uniform throughout the countries; rather, as expected, they are most concentrated within urban areas, often with an inter-tower spacing of less than 1 km. In rural areas with very sparse population density, the spacing amongst cell towers can be more than 100 km apart. This spatial inhomogeneity may pose concerns regarding the spatial accuracy of our home/work commuting characterization, especially in rural areas with large inter-tower spacing. While in CDR datasets, unlike GPS traces for example, typically the spatial resolution is beyond our control, we argue that commute is most interesting and relevant in urban and semi-urban areas, where cell towers in both Portugal and Ivory Coast are quite uniform and closely packed. For the rural commuters (for example, into/out of small provincial towns), the sparse spacing in rural areas may cause inaccuracies in two accounts: (1) it can grossly over-estimate the commute distances of people whose commute happens to cross from one cell tower to the next with a large inter-tower distance; and (2) it can grossly under-estimate the commute distances of people whose commute does not cross cell towers (whose home/work location would be identical). If it were possible to assume that everyone commutes, then on the aggregate population level, this error could be averaged out, giving rise to an estimate that approaches the true population mean. However, without this assumption, it became harder to accurate estimate rural commutes, without the aid of further datasets such as GPS traces on smartphones. Given the limited datasets available at our disposal, we did not undertake a detailed quantification of this effect, except noting that the people who undertake rural commuting make up a minority (in both countries, less than 5%) of the overall countrywide commuters. In our study, we filtered out users who spend significant time in a cell tower that is more than 50 km from adjacent ones, though in a future larger study with complementary datasets (such as smartphone GPS traces), there will be more elaborate measures that can be taken to ensure accuracy of this minority group of commuters.

On the other hand, while, as discussed above, GPS coordinates are generally more spatially accurate for quantifying human mobility compared to cell tower locations, it is still conceivable that in some areas (such as tunnels or under buildings), the ability to detect GPS signals may be impaired. If so, such locations may be under-represented in our data. However, we argue that these circumstances, in daily commuting conditions, are typically rare. As shown in Fig. S2(a), for reported GPS coordinates aggregated over one day in Milan, we can generally see clear delineation of roads, which is what we expect.

In the case of Saudi Arabia, we see that the distribution of commute distances again diverges significantly from Portugal and Boston under about 4 km. Similarly to Ivory Coast, many more individuals live closer to their places of work.

Finally, Milan represents a different case in which the dataset is GPS traces from cars instead of phone signaling data and therefor represents only a subsample of all the commuters - drivers. Here, while the initial distribution is qualitatively similar, the long tail falls off at a different slope, as shown in the inset. This is likely the effect of sub-selecting the mobility pattern in which individuals commute by cars. We see that in the long distance regime of above 30 km in commute distance, such commute distances are less frequent than the other aggregate mobility datasets. This may simply reflect the fact that it becomes less economical (from the perspective of time, fuel, and labor) to operate a car over long-distance commutes, in preference for other modes of transport (such as commuter trains) that may be available in the local context.

In order to better quantify the inherent differences in the timing of commute in different regions, we measured the peak commute times in each of the distributions above as a function of the commute distances. In order to calculate the peak commute times, we undertook two different methods. In the first method, we equated the peak commute time for each distribution to the median time from the entire distribution. This method would minimize the influence from extreme outliers (such as the low-level of activities as early in the morning). In the second method, we first fitted each distribution to a Gaussian distribution, and then equated the peak commute time with the mean of the fitted Gaussian. For each method, we plot results below in Figure 4 for morning commute (left column) and evening commute (right column), calculating the peak times using the median time method (first row) and the fitted Gaussian method (second row).

3a8082e126