Time Window Analysis Tutorial

This tutorial explores how to use time-based vulnerability analysis in VERUS.

Understanding Time Windows

Time windows capture how vulnerability changes throughout the day:

Businesses open and close at different hours
Population density fluctuates during the day
Emergency services may have different response times

Working with Time Windows

Create and customize time windows:

from verus.data import TimeWindowGenerator
import pandas as pd

# Create generator with a specific reference date
tw_gen = TimeWindowGenerator(
    reference_date="2023-11-06",  # A Monday
    verbose=True
)

# Generate standard time windows
time_windows = tw_gen.generate_from_schedule()

# View categories
print(f"Generated time windows for categories: {list(time_windows.keys())}")

# Examine a specific category
if "restaurant" in time_windows:
    print("\nRestaurant vulnerability throughout the day:")
    print(time_windows["restaurant"][["hour", "vi"]])

Creating Custom Time Windows

You can create custom time windows for specific scenarios:

# Create custom time window data
custom_tw = {}

# Define morning rush hour scenario (7-9 AM)
custom_tw["morning_rush"] = pd.DataFrame({
    "hour": range(24),
    "vi": [0.2] * 7 + [1.0] * 2 + [0.5] * 15  # High VI from 7-9 AM
})

# Define evening rush hour scenario (5-7 PM)
custom_tw["evening_rush"] = pd.DataFrame({
    "hour": range(24),
    "vi": [0.3] * 17 + [1.0] * 2 + [0.3] * 5  # High VI from 5-7 PM
})

# Define weekend scenario
custom_tw["weekend"] = pd.DataFrame({
    "hour": range(24),
    "vi": [0.1] * 10 + [0.7] * 8 + [0.3] * 6  # Medium VI from 10 AM-6 PM
})

Running Assessments with Different Time Windows

Compare vulnerability across different time scenarios:

from verus import VERUS
import pandas as pd

# Assuming you have poi_data and hex_grid ready

# Create empty placeholder for centroids
empty_centroids = pd.DataFrame(columns=["latitude", "longitude"])

# Initialize assessor
assessor = VERUS(
    place_name="Porto",
    method="KM-OPTICS",
    distance_method="gaussian"
)

# Load data once
assessor.load(
    potis_df=poi_data,
    centroids_df=empty_centroids,
    zones_gdf=hex_grid
)

# Run assessments for different time scenarios
scenarios = {
    "Morning Rush (7-9 AM)": "morning_rush",
    "Evening Rush (5-7 PM)": "evening_rush",
    "Weekend": "weekend"
}

results = {}

for scenario_name, eval_time in scenarios.items():
    print(f"\nEvaluating scenario: {scenario_name}")

    # Update evaluation time
    assessor.evaluation_time = eval_time

    # Run assessment with corresponding time window
    scenario_results = assessor.run(
        time_windows=custom_tw,
        area_boundary_path=None
    )

    # Store results
    results[scenario_name] = scenario_results

    if scenario_results["vulnerability_zones"] is not None:
        # Print summary statistics
        vz = scenario_results["vulnerability_zones"]
        print(f"Zones processed: {len(vz)}")
        print("Vulnerability statistics:")
        print(vz["VL_normalized"].describe())

        # Save results
        assessor.save(f"./results/{eval_time}/")

Comparing Time Window Scenarios

Visualize and compare vulnerability across scenarios:

import matplotlib.pyplot as plt
import seaborn as sns

# Prepare data for comparison
comparison_data = []

for scenario, res in results.items():
    if res["vulnerability_zones"] is not None:
        scenario_df = res["vulnerability_zones"][["VL_normalized"]].copy()
        scenario_df["scenario"] = scenario
        comparison_data.append(scenario_df)

# Combine data
all_scenarios = pd.concat(comparison_data)

# Create boxplot comparison
plt.figure(figsize=(10, 6))
sns.boxplot(x="scenario", y="VL_normalized", data=all_scenarios)
plt.title("Vulnerability Comparison Across Time Scenarios")
plt.ylabel("Normalized Vulnerability Level")
plt.xlabel("Time Scenario")
plt.xticks(rotation=45)
plt.tight_layout()
plt.show()

Conclusion

You’ve now learned how to use time-based vulnerability analysis in VERUS. This approach allows you to evaluate urban vulnerability under different temporal conditions, providing a more comprehensive understanding of risk patterns throughout the day.