This library combines land-use data (e.g. location of jobs, population, shops, healthcare, etc.) and pre-computed travel costs (e.g. travel times, transit fares, etc.) to generate transport accessibility metrics. Most of the work is conducted by manipulating pandas DataFrames. Current functionality allows for computing three types of accessibility measures. These include:
Potential accessibility measures: the sum of opportunities reachable from a location, weighted by their proximity (e.g. access to employment in a region, like how many jobs can be reached in a 45 minute commute)
Passive accessibility measures: the sum of the population who can access a location, weighted by their proximity (e.g. access to the labour force in a region, like how many workers can commute to a location within 30 minutes)
Minimum travel cost measures: the minimum travel cost to reach X opportunities (e.g. what is travel time to the nearest grocery store, or the minimum travel time to the nearest 3 libraries)
The library also includes functions for
estimating intra-zonal travel costs
filling in gaps in a travel cost matrix using a spatial weights matrices
generating travel impedance based on different functions (cumulative, linear, negative exponential, inverse power)
computing generalized costs
Planned future functionality will include competitive (i.e. floating catchment) measures of accessibility. Also on the to do list is to create proper documentation. For now, take a look at the basic usage and examples linked below.
pip install tracc
Requirements are pandas, numpy, geopandas, libpysal
# Loading in destination data.# For this example, these are job counts by block group from the the LEHD for Boston.dfo=tracc.supply(
supply_df=pd.read_csv("examples/test_data/boston/destination_employment_lehd.csv")
columns= ["block_group_id","C000"] # C000 pertains to the total number of jobs
)
# Loading in travel costs.# For this example, travel times by transit between block groups in Boston at 8am on June 30, 2020.dft=tracc.costs(
pd.read_csv(
"examples/test_data/boston/transit_time_matrix_8am_30_06_2020.zip",
compression='zip')
)
dft.data.time=dft.data.time/60# converting time from seconds to minutes# Computing impedance function based on a 45 minute travel time threshold.dft.impedence_calc(
cost_column="time",
impedence_func="cumulative",
impedence_func_params=45,
output_col_name="fCij_c45",
prune_output=False
)
# Setting up the accessibility object.# This includes joining the destination data to the travel time data.acc=tracc.accessibility(
travelcosts_df=dft.data,
supply_df=dfo.data,
travelcosts_ids= ["o_block","d_block"],
supply_ids="block_group_id"
)
# Computing accessibility to jobs based on the 45-min threshold.dfa=acc.potential(
opportunity="C000",
impedence="fCij_c45"
)Here's the top five rows of dfa (e.g. from block group 250056001001 someone can reach 4,061 jobs in a 45 minute transit trip)
o_block A_C000_fCij_c45
---------------------------------
0 250056001001 4061.0
1 250056001002 3960.0
2 250056002021 3608.0
3 250056002022 7845.0
4 250056002023 5124.0
This result can then be mapped in Python, QGIS, or any other mapping software by joining to the spatial data that pertain to these locations. Here's a quick example:
bg=gpd.read_file("example/test_data/boston/block_group_poly.geojson")
bg=bg.merge(dfa, left_on='GEOID', right_on="o_block", how="left")
bg.plot(column='A_C000_fCij_c45', figsize=(8, 8), scheme='quantiles', legend=True);