stevents

Lightweight S3 classes for spatio-temporal event pattern analysis in R

stevents provides tools for representing, manipulating, and analysing spatio-temporal point patterns. It was designed with wildfire ignition modelling in mind, but applies to any domain where events have both a location and a timestamp; seismology, epidemiology, criminology.

Installation

devtools::install_github("ricardosp4/stevents")

Core classes

Class	Description
stevents	A set of events with coordinates and timestamps
stgrid	A regular space-time tessellation for aggregating events

• stevents: in addition to print(), summary(), and plot(), it has subset() (filter by time/space) and as_sf() (convert to sf).
• stgrid: in addition to print(), summary(), and plot(), it has dim() (data-cube size) and as_sf() (convert cells to sf).

Functions

Function	Description
count_events()	Count events per grid cell and time bin
neighbor_history()	Count recent nearby events for each event (discrete excitation)
hawkes_intensity()	Compute the Hawkes self-excitation kernel (continuous)
as_sf()	Convert stevents to an sf point layer
subset()	Filter events by time window and/or bounding box
dim()	Get the size of an stgrid data cube (cells x time bins)

Quick start

library(stevents)

set.seed(1)
ev <- stevents(
  x    = runif(200, 0, 100),
  y    = runif(200, 0, 100),
  time = as.POSIXct("2023-01-01") + sort(runif(200, 0, 365 * 86400))
)

print(ev)
#> <stevents> spatio-temporal event pattern
#>   Events    : 200
#>   Time range: 2023-01-01 16:05:27 -> 2023-12-30 12:39:27
#>   Bounding box:
#>     x: [1.30776, 99.2684]
#>     y: [2.77871, 99.6077]
#>   CRS       : NA
summary(ev)
#> Summary of <stevents>
#>   Number of events     : 200
#>   Time span (days)     : 362.86
#>   Mean events per day  : 0.5512
#>   Bounding-box area    : 9485
#>   Mean events per area : 0.02108

plot(ev)

Hawkes kernel

ev2 <- hawkes_intensity(ev,
  alpha = 0.8,
  beta  = 1 / (7 * 86400),
  gamma = 1 / 30
)

summary(ev2$data$hawkes_intensity)
#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#>  0.0000  0.4780  0.6917  0.7479  0.9393  2.0585

The continuous kernel captures finer variation than a fixed binary window, events outside a discrete neighbourhood can still carry non-trivial excitation.

ev3 <- neighbor_history(ev, radius = 15, lag = 30 * 86400)
ev3 <- hawkes_intensity(ev3, alpha = 0.8,
                        beta  = 1 / (7 * 86400),
                        gamma = 1 / 30)

plot(
  ev3$data$n_recent_neighbors,
  ev3$data$hawkes_intensity,
  xlab = "n_recent_neighbors (discrete)",
  ylab = "Hawkes intensity (continuous)",
  pch  = 16,
  col  = adjustcolor("#E05A2B", 0.4),
  main = "Discrete vs continuous excitation"
)

Included datasets

Dataset	Description	Source
venezuela_2026	Seismic sequence following the Mw 7.5 earthquake of 24 June 2026	USGS
amatrice_2016	Amatrice–Norcia aftershock sequence, August–November 2016 (74 events)	USGS

data(amatrice_2016)

events_italy <- stevents(
  x    = amatrice_2016$lon,
  y    = amatrice_2016$lat,
  time = amatrice_2016$time
)

print(events_italy)
#> <stevents> spatio-temporal event pattern
#>   Events    : 74
#>   Time range: 2016-08-24 01:36:32 -> 2016-11-14 19:49:52
#>   Bounding box:
#>     x: [12.9588, 13.3022]
#>     y: [42.6, 43.1228]
#>   CRS       : NA

Motivation

stevents was developed as part of a Master’s thesis on spatio-temporal Hawkes process modelling of wildfire ignition risk on the Iberian Peninsula. The package provides the preprocessing layer between raw event data and intensity model fitting via INLA.

License

Copyright (c) 2026 Ricardo Sales Piquer