Introduction
Hot spot analysis is a statistical tool to identify clusters in a dataset. In traffic management, identifying hot spots of
This article presents an exemplary hot spot analysis in the city of Hamm using data collected in calendar weeks 26 to 34 (June-August) in 2020. The data is based on a very small subset (about 900 trajectories) with a biased selection of drivers (mainly employees driving staff cars) during typical working hours. As a result, the following analysis does not necessarily represent the actual traffic situation in Hamm, but discusses the general insights that a hot spot analysis can generate.
Methodology
Getis Ord statistics
Spatial association is often analyzed using G statistics [1, 2, 3]. We first divide the area of interest into
where the star (
Under Gaussianity assumptions,
A significantly large
Analysis steps
We used Jupyter Notebooks and the Python libraries envirocar-py, pandas, geopandas, numpy, pysal (esda, libpysal) to do the analysis and applied the following steps to obtain hot spot maps.
- Read raw xFCD; extract coordinates,
emissions in kg/h and vehicle speed in km/h; optionally normalize emissions by vehicle speed to emissions in kg/km - Remove outliers
- Aggregate
emissions on equidistant grid with 30 m spacing - Calculate mean values of
emissions for each grid cell - Calculate row-standardized binary weights with a fixed distance band of 300 m (10 cells)
- Calculate
statistics for each feature (i.e. grid cell) - Under Gaussianity assumptions, define cells that are within the upper/lower 10 % of all
-values as hot spot or cold spot, respectively
Results
In this section, we present one hot spot map for the original
Figure 1 shows a hot spot map for the original

Cars might emit less

The representation of the data strongly influences the detection of hot and cold spots as you can see from Figure 1 (based on emissions in kg/h) and Figure 2 (based on emissions in kg/km). There are also various ways to tune and tweak the hot spot analysis: e.g. which metric
Another shortcoming might be the strict discretization in merely cold spots, grey spots and hot spots. The transition between these might also shed some light on the underlying features. As an example (Figure 3), we grouped the

Conclusion
We used xFCD from the enviroCar platform to perform a hot spot analysis of
The analysis performed represents an exploratory view on the data and can reveal general patterns. In order to answer specific questions, we need to adapt the algorithm, e.g. by performing the analysis on specific street segments or selecting specific time windows.
The CITRAM project is funded by the German Federal Ministry of Transport and Digital Infrastructure within the mFund framework under FKZ 19F2068.
Authors: Martin Pontius, Benedikt Gräler, Albert Remke
References
[1] Getis, A., & Ord, J. K. (1992). The analysis of spatial association by use of distance statistics.
[2] Ord, J. K., & Getis, A. (1995). Local spatial autocorrelation statistics: distributional issues and an application. Geographical analysis, 27(4), 286-306.
[3] Anselin, L. (1995). Local indicators of spatial association—LISA. Geographical analysis, 27(2), 93-115.
Leave a Reply