Uganda wildfires – A Bayesian approach for forecasting VIIRS-SNPP fires on a discrete global grid
M.G. Walsh
12, March, 2024
1 Introduction
Fires are a natural and beneficial process in landscapes. Fires recycle nutrients into the soil, and play an essential role in primary and secondary succession of plant communities in rangelands, forests and wetlands / peatlands. In managed rangelands and forests prescribed burns are used for woody vegetation as well as understory vegetation control. However, when vegetation burns, it emits greenhouse gases (GHGs) and various other materials into the atmosphere. Wildfires can also be extremely destructive when they impinge on urban areas. In many regions around the world the increasing frequency and severity of wildfires, exacerbated by climate change, have now underscored the need for near-real-time spatial fire forecasting for land management and disaster response.
In this notebook, I develop a predictive Bayesian repeated measures approach on a Discrete Global Grid (DGG). The approach utilizes openly available fire detection time series from the Visible Infrared Imaging Radiometer Suite - Suomi National Polar-orbiting Partnership (VIIRS-SNPP) satellite for Uganda (2012 – 2023). VIIRS-SNPP, which operates globally on all fire affected areas, is equipped with multiple spectral bands that are sensitive to the wavelengths of light emitted by fires. Specifically, it uses the middle-infrared (MIR) and thermal-infrared (TIR) bands to detect the heat signature of fires. The MIR band is particularly useful for identifying hotspots that indicate active fires, while the TIR band helps in estimating the temperature of the detected fires.
The goal is to uncover spatial patterns and temporal trends that are often overlooked. By doing so, the associated models can forecast potential hotspots in Uganda for the upcoming year. They can also be used for fire risk mapping and management. This data-driven approach could enable more strategic efforts and use cases to mitigate wildfire risks and impacts while harnessing their ecological benefits, and provide a basis for better-informed land management policies. The aim is not just to react to fires; it’s about anticipating them. Such a proactive stance in fire management could significantly reduce the associated risks and their environmental impacts.
This R markdown document is maintained and versioned on the Uganda OSF repository here, from where it can be downloaded and modified. Please cite this work as: M.G. Walsh (2024). Uganda wildfires – A Bayesian approach for forecasting VIIRS-SNPP fires in a space-time cube. https://doi.org/10.17605/OSF.IO/4RMKP.
2 Example use cases
2.1 Prescribed burns for vegetation and soil management
Prescribed burns, also known as controlled burns, are an essential tool in managing vegetation, reducing wildfire risks, and enhancing soil fertility. Utilizing the wildfire forecasting system, land management agencies can identify optimal times and locations for prescribed burns that minimize risk and maximize ecological benefits. Controlled fires help to clear underbrush, dead trees, and other flammable materials, reducing the fuel available for wildfires. They also play a crucial role in maintaining ecosystem health by promoting the growth of native plant species and improving the availability of soil nutrients and reducing soil acidity where it is present.
The forecasting system assesses various factors, including historical patterns such as those used in this notebook, but also including, ground survey investigations, weather conditions, vegetation types, and fire behavior models, to determine the ideal periods for conducting prescribed burns. This data-driven approach ensures that burns are conducted safely, with minimal impact on air quality and surrounding areas. The process involves:
Planning: Identifying target areas for vegetation management based on fire risk assessments and ecological goals.
Timing: Selecting the optimal time for burns, considering weather patterns, humidity levels, and wind conditions to control the burn effectively and safely.
Execution: Implementing the prescribed burn under controlled conditions with prepared firefighting resources on standby to manage any unforeseen spread of fire.
Monitoring: Continuously monitoring the burn area and surrounding regions for any signs of unintended fire spread and assessing the effectiveness and ecological impact of the burn.
Post-burn analysis: Evaluating the impact on vegetation control, soil fertility enhancement, and the reduction in wildfire risk to inform future prescribed burn plans and adjust strategies as necessary.
Incorporating prescribed burns into wildfire management strategies, guided by precise forecasting and environmental data, ensures that these activities are both beneficial for land management and instrumental in mitigating the threat of uncontrolled wildfires. This proactive approach seeks to support biodiversity, aids in the restoration of natural fire regimes, and contributes to the health and resilience of forests, rangelands and wetlands/peatlands.
2.2 Integrated wildfire management system for emergency response and public health
This integrated approach combines early warning systems, firefighting strategy optimization, and public health advisories into a comprehensive wildfire management system. Leveraging advanced forecasting technologies, the system enhances emergency preparedness, directs firefighting resources efficiently, and minimizes public health risks from wildfire smoke. By predicting wildfire outbreaks and behavior, it enables timely evacuation warnings, strategic deployment of firefighting resources, and proactive public health measures to safeguard communities and ecosystems including:
Early warning and evacuation planning: The system uses predictive modeling to identify areas at high risk of wildfires, issuing early warnings to those regions. It aids in devising evacuation plans, ensuring safety protocols are in place for vulnerable communities.
Resource allocation and firefighting strategy: Based on forecasted fire behavior and spread patterns, the system allocates firefighting resources strategically. It prioritizes areas for intervention, optimizes the use of aerial and ground firefighting assets, and identifies safe and effective firefighting lines.
Public health advisory: By predicting the direction and dispersion of wildfire smoke, the system issues health advisories to affected communities, recommending precautions to reduce exposure to smoke and safeguarding public health. It also guides the implementation of air quality management strategies in real-time.
Operational execution: Operational teams use the system’s insights to execute evacuation plans, firefighting operations, and public health advisories. Continuous monitoring and real-time data feed into the system, allowing for dynamic adjustments to strategies as conditions evolve.
Impact assessment and feedback loop: Post-event analysis evaluates the effectiveness of the response strategies, the accuracy of forecasts, and the health outcomes. This feedback is used to refine models, improve prediction accuracy, and enhance response strategies for future incidents.
This use case emphasizes a coordinated and data-driven approach to managing wildfires, integrating emergency response, firefighting efforts, and public health considerations. By leveraging advanced forecasting and analytics, the system ensures that responses are timely, resources are utilized efficiently, and communities are protected against the myriad risks posed by wildfires.
2.3 Wildfire urban interface management system
The Wildfire Urban Interface (WUI) Management System is designed to specifically address the unique challenges of managing wildfires in areas where urban development meets or intermingles with wildland or vegetative fuels. This system focuses on reducing the vulnerability of these areas through strategic planning, community engagement, and the integration of fire-adapted building and landscaping practices. Utilizing predictive analytics and risk assessment tools, the system aids in identifying high-risk zones, optimizing mitigation efforts, and ensuring rapid response capabilities to protect lives, properties, and ecosystems. Implementing a comprehensive approach involves collaboration between fire management agencies, urban planners, and the community to develop and enforce strategies tailored to the WUI’s unique conditions:
Risk assessment and mapping: Employing advanced spatial analysis and modeling to identify high-risk areas within the WUI, considering factors such as vegetation type, topography, climate conditions, and urban density. This assessment informs the development of detailed risk maps that guide mitigation and response efforts.
Mitigation strategies: Based on risk assessments, the system facilitates the planning and implementation of targeted mitigation strategies, such as creating defensible spaces around properties, enforcing fire-resistant building codes, and managing vegetation to reduce fuel loads.
Community engagement and education: Developing and deploying educational programs aimed at residents and property owners in the WUI to promote awareness of wildfire risks and encourage proactive measures to reduce vulnerability, such as adopting fire-smart landscaping practices and participating in community fire preparedness programs.
Emergency planning and response coordination: Integrating real-time data from the forecasting system to support emergency planning, including evacuation routes, shelter locations, and communication plans. Enhancing coordination among firefighting units, emergency services, and community response teams to ensure rapid and effective response to wildfires.
Recovery and resilience building: Post-fire assessments to evaluate the effectiveness of mitigation measures and response strategies, with a focus on learning and adaptation. Initiatives to rebuild affected areas with enhanced resilience, incorporating fire-adapted designs and materials in reconstruction efforts, and restoring natural landscapes to reduce future wildfire risks.
The Wildfire Urban Interface Management System represents a holistic approach to reducing the impact of wildfires in vulnerable urban-edge communities. By combining predictive analytics with strategic planning and community engagement, this system aims to create more resilient WUI areas that can withstand and recover from wildfire events, thereby safeguarding human life and property.
2.4 Climate Change Adaptation for Rural Communities
This use case focuses on leveraging the wildfire forecasting system to aid rural communities in adapting to the challenges posed by climate change. As these areas often face heightened risks of wildfires due to changing climate conditions, the system provides insights for long-term planning and resilience building. It assesses how shifting weather patterns, increased temperatures, and altered precipitation levels contribute to wildfire risks, offering rural communities evidence-based strategies for adaptation and mitigation. The adaptation process involves a series of coordinated steps, integrating the forecasting system’s capabilities with community planning and environmental management practices:
Vulnerability assessment: Utilizing the forecasting system to analyze historical and predictive data on climate change impacts specific to rural regions. This includes identifying areas most at risk of experiencing increased wildfire frequency and intensity.
Adaptation strategy development: Based on the vulnerability assessment, developing comprehensive adaptation strategies that address both immediate and long-term needs. This may include altering land use practices, enhancing vegetation management to reduce fuel loads, and implementing water conservation measures to mitigate drought conditions.
Community capacity building: Engaging with local communities through workshops, training sessions, and educational programs to raise awareness about climate change impacts on wildfire risks. Empowering residents with the knowledge and tools to participate in mitigation efforts, such as creating defensible spaces and adopting fire-resistant building materials and techniques.
Emergency response planning: Incorporating climate-adapted wildfire forecasts into emergency preparedness plans, ensuring that rural communities have updated evacuation plans, emergency supplies, and communication systems in place to respond effectively to wildfire threats.
Monitoring: Establishing a feedback loop where ongoing climate and wildfire risk assessments inform adjustments to adaptation strategies, ensuring they remain effective as conditions change.
This use case emphasizes the role of advanced fire forecasting in supporting rural communities’ efforts to adapt to climate change. By providing actionable intelligence on future wildfire risks, the system enables these communities to implement targeted adaptation measures, build resilience, and reduce their vulnerability to the increasing threats posed by wildfires in a changing climate.
3 Setup
To run this notebook, you will need to install and load the R-packages indicated in the chunk directly below.
# Package names
packages <- c("osfr", "tidyverse", "patchwork", "rgdal", "sf", "raster",
"zoo", "brms", "mgcv", "ggridges", "DT")
# Install packages
installed_packages <- packages %in% rownames(installed.packages())
if (any(installed_packages == FALSE)) {
install.packages(packages[!installed_packages])
}
# Load packages
invisible(lapply(packages, library, character.only = TRUE))3.1 Read VIIRS fire detections
The archived VIIRS-SNNP fire data are openly available at FIRMS. Do note
the you have to place a request for the newer (NRT) data for 2022 –
2023. You can get more information at what level those data have been
processed to from the FIRMS FAQ.
I have put a clean, concatenated file in the Uganda OSF repository here, and you can download it manually
from there. You can also use the following
osfr download option (recommended):
# download VIIRS data from OSF
osf_retrieve_node("np3wy") %>%
osf_ls_files(n_max = Inf) %>%
osf_download(path = "./fires", conflicts = "overwrite")viirs <- read.table("./fires/viirs_ts.csv", header = TRUE, sep = ",")The following is an animation of the VIIRS fire detection data for the years 2012 – 2023 on a monthly time step. Each VIIRS active fire/thermal hotspot location represents the center of a 375 meter pixel. Wildfires are commonplace in Uganda, and the distinctive space-time oscillation pattern that repeats every year, synchronizes with the latitudinal movement of the Intertropial Convergence Zone (ITCZ) near the equator.
