locationallocation (v0.1)

locationallocation: an R package to solve Maximal Coverage Location-Allocation problems using geospatial data

Package developer: Giacomo Falchetta, giacomo.falchetta@cmcc.it

• Documentation and vignette website: https://giacfalk.github.io/locationallocation

• Reference preprint (under review at a peer-reviewed journal): https://doi.org/10.31223/X5XQ69

@article{Falchetta2025,
  title = {locationallocation: solving Maximal Coverage Location-Allocation geospatial infrastructure assessment and planning problems},
  url = {http://dx.doi.org/10.31223/X5XQ69},
  DOI = {10.31223/x5xq69},
  publisher = {EarthArXiv},
  author = {Falchetta,  Giacomo},
  year = {2025},
  month = mar 
}

Background

Assessing and planning infrastructure and networks over space conditional to a spatially distributed demand and with consideration of accessibility and spatial justice goals and under infrastructure allocation constraints is a key policy objective. This class of problems is generally defined as "Maximal Coverage Location-Allocation (MCLA)" spatial optimisation problems.

locationallocation, an R package to solve MCLA problems using geospatial data in widely used R programming language geospatial libraries. The locationallocation package allows to produce travel time maps and spatially optimise the allocation of facilities/infrastructure based on spatial accessibility criteria weighted by one or more variables or a function of those.

Potential applications

Potential applications of the package extend to the domains of public infrastructure assessment and planning (public services provision, e.g. transport, social services, healthcare, parks), urban environmental and climate risk reduction interventions, logistics and hubs allocation, commercial and strategic decisions.

Installation

Install with:

library(devtools)
install_github("https://github.com/giacfalk/locationallocation")

Operation

Operate the package as follows.

First, load the package.

library(locationallocation)

As an example, We demonstrate how the package can tackle urban-scale climate risk through robust infrastructure assessment and geospatial planning. We run the demo_data_load() function to load a set of demo datasets to run the package's function. The demo data contains the coordinate-point location of public drinking water fountains in the city of Naples, Italy, as well as a gridded population raster data from GHS-POP, a 100-m resolution map of heat hazard (number of days with Wet-Bulb Globe Temperature greater than 25° C in the historical period 2008-2017, obtained from the UrbClim model), and the administrative boundaries of the city.

Then, we can use the traveltime function to generate a map of the current accessibility to the facility point sf input (here, fountains) within the specified geographical boundaries, and with a choosen travel mode (walk or fastest), and a given output spatial resolution (in meters; achieved using disseving spatial downscaling techniques). Once the function has run successfully, we can generate a map of the resulting layer using the traveltime_plot function.

demo_data_load()

out_tt <- traveltime(facilities=fountains, bb_area=boundary, dowscaling_model_type="lm", mode="walk", res_output=100)

traveltime_plot(traveltime=out_tt,  bb_area=boundary, facilities = fountains)

We can also produce a summary plot and statistic based on the output of the traveltime function and a given demand (e.g., population) raster, as well as a given time threshold parameter:

traveltime_stats(traveltime = out_tt, demand_raster = pop, breaks=c(5, 10, 15, 30), objectiveminutes=5)

which will print a summary cumulative curve plot and an output message such as:

[1] "38.54 % of demand layer within the objectiveminutes threshold."

We can now use the allocation function to optmise the spatial allocation of new water fountains to ensure that (virtually) everyone (i.e., the totality of the raster layer specified by the demand_raster parameter) can walk to one within 15 minutes, as specified by the objectiveminutes parameter:

output_allocation <- allocation(demand_raster = pop, traveltime_raster=out_tt, bb_area = boundary, facilities=fountains, weights=NULL, objectiveminutes=15, objectiveshare=0.99, heur="max", dowscaling_model_type="lm", mode="walk", res_output=100)

allocation_plot(output_allocation, bb_area = boundary)

Note that is is also possible to solve an allocation problem specifying a weights parameter, to attribute more relative importance or priority to areas where the demand_raster is overlapping with some weighting factors (defined by another raster layer), such as exposure to hot days, as in the following example:

output_allocation_weighted <- allocation(demand_raster = pop, traveltime_raster=out_tt, bb_area = boundary, facilities=fountains, weights=hotdays, objectiveminutes=15, objectiveshare=0.99, heur="max", dowscaling_model_type="lm", mode="walk", res_output=100)

allocation_plot(output_allocation_weighted, bb_area = boundary)

A variant of the allocation problem is the case when the set of candidate locations to allocate new faciltiies is discrete (and not continuous over the study area, as in the previous example). In this case, the user needs to provide a discrete set of location points into the candidate parameter of the allocation_discrete function, as well as a maximum number of facilities (n_fac parameter) that can be selected among the candidate locations. The function will apply a quasi-optimality heuristic (using a randomisation based approach, where the number of replications - defined by the n_samples parameters - will gradually approach the global optimum but it will linearly increase the computational time.

candidates <- st_sample(boundary, 30)

output_allocation_discrete <- allocation_discrete(demand_raster = pop, traveltime_raster=NULL, bb_area = boundary, facilities=fountains, candidate=candidates, n_fac = 10, weights=NULL, objectiveminutes=15, dowscaling_model_type="lm", mode="walk", res_output=100, n_samples=100)

allocation_plot_discrete(output_allocation_discrete, bb_area = boundary)

Consider also the case of a problem where there are no existing facilities to start with, and hence the discrete location-allocation problem needs to optimise the allocation to cover as much as possible of the demand within the limited set of discrete choices over space to allocate new facilities, as well as the constraint set by the maximum number of allocable facilities:

set.seed(333)

output_allocation_discrete_from_scratch <- allocation_discrete(demand_raster = pop, traveltime_raster=NULL, bb_area = boundary, facilities=NULL, candidate=candidates, n_fac = 10, weights=NULL, objectiveminutes=15, dowscaling_model_type="lm", mode="walk", res_output=100, n_samples=100)

allocation_plot_discrete(output_allocation_discrete_from_scratch, bb_area = boundary)

Disclaimer

This package is developed by an R user without a professional supervision or a testing environment or protocol. As such, the developer does not hold any responsibility for the results produced by this package.