-- MATSim Open Berlin Scenario --

Properties:
* Full-day plans
* 3 modes: car, pt, slowPt
* stable in terms of route and mode choice

The approach:
* The population for the scenario covers all Berlin and Brandenburg and has been generated based on the “Zensus 2011” (*).
* Workplaces for employees are informed by the “Pendlerstatistik 2009” on a municipality level. Note that Berlin is only one municipality, which makes this approach requiring refinement inside Berlin (while it is well-suited as it is for areas with smaller municipalities like mainly found in Brandenburg) (*).
* In terms of the choice of workplace locations for people in Berlin as well as secondary locations, the approach works as described by Ziemke, Nagel, Bhat (2015) (WP 14-15 on https://www.vsp.tu-berlin.de/publications/vspwp/). Very briefly: Daily activity-travel plans are generated for every agent four times, such that each agent has a set of four potential plans which differ in terms of activity locations (next to additional minor variations concerning other choice criteria).
* Based on some 8,000+ traffic counts (i.e. 300+ traffic count stations with each 24 hourly counts), a Cadyts calibration has been used to choose those plans from each agents plan set that best contributes to a traffic pattern that resembles reality (Reality as given by mentioned traffic counts). (*) (***)
* After each agent has been assigned with locations by the previous step (**), a mode choice calibration has been done which separates all agents into the car and two different teleported (pt) modes.
- “Car" with a monetary distance rate of -0.0002; default parameters otherwise
- “Pt” with a constant of -6.0 and a speed of 11.111m/s; default parameters otherwise
- “SlowPt” with a constant of -3.0 and a speed of 4.167m/s; default parameters otherwise
* Note that in the previous step (responsible for mode choice), no (Cadyts) calibration is active. As such, this step can also be regarded as a stability test in the sense of Ziemke, Nagel, Bhat (2015), i.e. a simulation run that confirms that choices stay that were initially made under the influence of Cadyts stay roughly the same after this influence is removed.

(*) = More details on this seem out of scope here; they can be found in a the documentation (/shared-svn/studies/countries/de/berlin_scenario_2016/documentation.pdf) that I will be updating soon.
(**) = You may have wondered why activities for *all* agents can be chosen in this step already, where the selection of modes is not yet observed. This is taken care for. Explaining seems out of scope here. Talk to me about it or check in the documentation soon.
(***) = The result of this step, if you are interested, is under "runs-svn/berlin_scenario_2016/be_115a"

Some results (validation):
* Altogether, there are 16.42m trips.
* Out of these, 3.17m trips are done by car and are Berlin-based and <100km long. The reference value from SrV 2008 is 3.20m trips.
* The average travel time for these trips is 23.1min, the distance 11.1km (references from SrV 2008: 22.3km, 9.5km, respectively).
* More importantly, see the plots below, where red = this scenario and blue = reference from SrV2008 below… (more plots in the run directory)
* Also see the Google Earth screenshot below, which shows a count-simulation comparison for the morning peak.