It's All about Streets Continued from page 47 Enhancing Street Data for Networks Here comes the really important stuff! When you are sourcing (or have already obtained) a street dataset, what important network issues should you address? In a recent ArcUser article, "Do It Yourself—Building a network dataset from local agency data" (referenced above), I introduced several important issues you should consider when preparing a dataset for network modeling. Here are key issues that can be tackled by spatial and attribute editing. Impedance—Distance To properly calculate travel time, you need accurate distances. In a well-defined and documented model, distance-based impedance is easy to calculate. If your streets are registered in a standard projected coordinate system, you can readily calculate and verify distances between intersections in the native coordinates. If you are careful, you can calculate distance in a geographic data frame or across the metric/ imperial measurement divide. I use a field named Length_Mi to store travel distance. Since this is not a reserved field name, you must specify this field and its units when you build a network. I do not use the field name Length, as I often store length in multiple units of measurements (both metric and imperial) and certainly do not want to confuse these systems. As you edit your network dataset, recalculate segment lengths frequently, especially before you rebuild a network. Impedance—Speed A reliable time-based impedance (or speed) field is essential. Many agencies start with posted or ordinance speeds and factor them up or down, depending on policy, persistent travel constraints (such as slope and turn radius), and variable travel factors (weather, traffic, and disruptions, to name a few). Measured travel times allow a user to fine-tune speed-based impedance. Once directionality and connectivity issues are resolved, the segment length in miles and travel time in minutes can be calculated. If you use a field named Minutes, Network Analyst will recognize it immediately. Remember to recalculate segment length, then travel times ArcUser Fall 2009 on all street segments (or at least the ones edited) each time you finalize an edit session and before you rebuild your network. Be sure to edit or remove all zero-time segments in the dataset before you rebuild. Crossing Relationships Overpasses and intentionally nonconnecting crossings are two of the most difficult and important issues to address. Commercial streets use an integer code or Z elevation/Z level to manage crossing geometries. The Z elevation fields (from and to) specify whether traffic may turn at an endpoint intersection. To turn or traverse through an intersection, the endpoint of the traveled segment must match the start point of the new segment. Most simple networks will use only Z values of 0 and 1. However, a Los Angeles County commercial dataset that I recently modeled included Z elevation values up to 4. If you add new streets to a dataset that uses Z elevations, be sure to correctly populate this field. A simpler way to manage crossing relationships, especially streets not also used for geocoding, is to build nonintersecting crossing geometries and apply endpoint connectivity. These streets are much easier to edit and visually inspect, and they perform much better if you are using lidar terrain to define slope impedance. Connectivity Connectivity defines how an individual element in a network relates to its neighbors. Simple connectivity options include endpoint and any vertex. I strongly recommend that you use endpoint connections. See Crossing Relationships, in the previous section, for more information. Simple single-layer networks must have exact endpoint connections to function properly. I find that nonconnecting segments are the most common issue that will cause a network to fail. Use visual inspection, polyline topology, editing, and/or old-fashioned trial and error to define, repair, and maintain connectivity. If you use a multimodal network that includes several feature classes, connectivity between participants is defined with special rules. Remember that you must place all participating feature classes in a single feature dataset to build multimodal models. www.esri.com