coventry_historical_report.qmd

---
title: "Historical Analysis of Coventry's Cycle Network (2017-2026)"
author: "Robin Lovelace"
date: "2026-02-23"
format:
  html:
    self-contained: true
    code-fold: true
    toc: true
    theme: cosmo
---

```{r setup, include=FALSE}
library(sf)
library(dplyr)
library(ggplot2)
library(osmactive)
library(tidyr)

# Load the saved data
if (!file.exists("coventry-cycleways-2017-2026.gpkg")) {
  stop("Data file not found. Please run the data generation script first.")
}
cycle_data = st_read("coventry-cycleways-2017-2026.gpkg", quiet = TRUE)

# GeoPackage might have converted factors to characters
# Re-apply factor levels from get_palette_npt() keys or osmactive standards
infra_levels = c(
  "Segregated Track (wide)",
  "Off Road Path",
  "Segregated Track (narrow)",
  "Shared Footway",
  "Painted Cycle Lane"
)
cycle_data$cycle_segregation = factor(cycle_data$cycle_segregation, levels = infra_levels)
```

# Introduction

This report provides a detailed analysis of the growth of cycling infrastructure in Coventry from 2017 to 2026. We compare a simple metric based solely on the `highway=cycleway` OpenStreetMap (OSM) tag with a more sophisticated classification system provided by the `osmactive` R package.

# Detailed Infrastructure Growth

The following plot shows the cumulative growth of the Coventry cycle network, broken down by the `osmactive` infrastructure types. We use the ordered factors returned by the package to ensure that high-quality infrastructure is prioritized in the visualization.

```{r stacked-bar}
# Prepare data for stacked bar chart: cumulative length per year per type
years = 2017:2026

# Calculate cumulative lengths for each type in each year
stacked_data_list = lapply(years, function(yr) {
  # Get all segments constructed in or before the current year
  sub_year = cycle_data %>% filter(year <= yr)
  
  if (nrow(sub_year) == 0) return(NULL)
  
  # Calculate lengths first
  sub_year$seg_length = as.numeric(st_length(sub_year))
  
  # Group by type and calculate length
  sub_year %>%
    st_drop_geometry() %>%
    group_by(cycle_segregation) %>%
    summarise(
      length_km = sum(seg_length, na.rm = TRUE) / 1000,
      .groups = "drop"
    ) %>%
    mutate(year = yr)
})

# Combine all years and ensure we have all levels represented for consistent plotting
stacked_data = bind_rows(stacked_data_list) %>%
  complete(year, cycle_segregation, fill = list(length_km = 0))

# Ensure factor order
stacked_data$cycle_segregation = factor(stacked_data$cycle_segregation, levels = rev(infra_levels))

ggplot(stacked_data, aes(x = factor(year), y = length_km, fill = cycle_segregation)) +
  geom_bar(stat = "identity", position = "stack") +
  # We want the legend and stack to be: Wide at top, then Path, then Narrow, then Shared, then Painted at bottom
  scale_fill_manual(values = rev(get_palette_npt()), name = "Infrastructure Type") +
  labs(
    title = "Coventry Cycle Network Growth by Type",
    subtitle = "Cumulative growth using osmactive classification (2017-2026)",
    y = "Total Length (km)",
    x = "Year"
  ) +
  theme_minimal() +
  theme(
    legend.position = "right",
    axis.text.x = element_text(angle = 45, vjust = 1, hjust = 1)
  )
```

## Animated Growth Map

The expansion of the network can be seen spatially in the animation below:

![Animation showing Coventry Cycle Network Growth from 2017 to 2026.](coventry_historical_growth.gif)

# Simple Metric: `highway=cycleway`

A common but limited approach to quantifying cycling infrastructure is to filter for segments where the `highway` tag is explicitly set to `cycleway`.

```{r simple-metric}
# Calculate cumulative growth for highway=cycleway
simple_growth = lapply(2017:2026, function(yr) {
  sub = cycle_data %>% 
    filter(year <= yr, highway == "cycleway")
  if (nrow(sub) == 0) return(data.frame(year = yr, total_length_km = 0))
  len = as.numeric(sum(st_length(sub))) / 1000
  data.frame(year = yr, total_length_km = len)
}) %>% bind_rows()

ggplot(simple_growth, aes(x = year, y = total_length_km)) +
  geom_line(color = "blue", linewidth = 1) +
  geom_point(color = "red", size = 2) +
  labs(
    title = "Growth of 'highway=cycleway' in Coventry",
    subtitle = "Simple tag-based metric (2017-2026)",
    y = "Total Length (km)",
    x = "Year"
  ) +
  theme_minimal()
```

### Limitations of the Simple Approach

1.  **Underestimation**: Many cycle paths are tagged as `highway=path`, `highway=footway`, or even `highway=service` with additional tags like `bicycle=designated`.
2.  **Overestimation**: Some segments tagged `highway=cycleway` may be substandard (e.g., extremely narrow or poorly maintained) but are treated the same as high-quality segregated tracks.
3.  **Missing Context**: This metric fails to distinguish between segregated tracks, shared-use paths, and painted lanes on the road.

# The `osmactive` Classification System

The `osmactive` package uses a multi-layered classification logic that is currently used in **production environments** for active travel planning and auditing. Instead of relying on a single tag, it analyzes a combination of:
- `highway` type
- `segregated` status
- `bicycle` and `foot` access tags
- Proximity to the driving network (via `distance_to_road()`)
- Physical dimensions (via `clean_widths()`)

This system classifies infrastructure into meaningful levels of service, ranging from "Segregated Track (wide)" to "Painted Cycle Lane".

# Conclusion

The analysis reveals a significant expansion of the cycle network in Coventry, particularly in the last few years. While the simple `highway=cycleway` metric shows growth, the `osmactive` classification provides a much clearer picture of the *quality* of the infrastructure being delivered, highlighting a shift towards more segregated and shared-use provision.

---
*Report generated on `r Sys.Date()` using historical OSM snapshots from the ohsome API.*