As the climate crisis continues to escalate, the frequency of climate hazards such as floods, landslides and wildfires is increasing. The result is increased death and injury, environmental damage and economic loss. Disaster risk reduction (DRR) systems aim to reduce disaster risk by better monitoring the risk of hazards, building capacity to detect and respond as quickly as possible, and improving community preparedness.
The frequency of natural disasters is increasing despite these DRR efforts. In 2023 alone, the Emergency Events Database recorded 399 natural disasters, 39 more than the 2003 to 2022 annual average. These events are estimated to have impacted more than 93 million people, resulting in almost 86,500 deaths and over $202 billion in economic damage.
Artificial intelligence (AI) presents an opportunity to improve the effectiveness of traditional DRR systems. Start-ups and Big Tech are offering a range of AI-enabled DRR solutions, which focus on a combination of predicting hazards, detecting disasters as early as possible, and responding to them more effectively. For example, in 2022, Google launched the Flood Hub, a free-to-use global platform which uses AI models to provide flood risk detection seven days in advance. Academic research centres across the globe are also developing AI models capable of predicting natural disasters such as landslides and earthquakes, generating large-scale building damage assessments and providing early warnings for tsunamis, to name a few.
Combatting forest fires with AI in Pakistan
GSMA Mobile for Development recently conducted research exploring the potential of AI for DRR, specifically in forest fire management in Pakistan, a country ranked 5th in the climate risk index globally. AI-enabled forest fire management solutions are proliferating in both high-income countries prone to risk, such as the US, Australia, Italy, Spain and Turkey, and in low and middle-income countries such as Ethiopia and Nepal, though in the latter context, limited resources can hinder their scalability.
Implementing an AI-enabled forest fire management system in Pakistan could relieve the strain on its overstretched forestry departments and enhance their resource management capabilities by mitigating the risk from natural hazards more effectively. AI can help predict the risk of a fire event, or determine the spread and intensity of a forest fire, by leveraging data on key factors that impact forest fire outbreaks, including the type and condition of vegetation in the forest, weather conditions, topography, and human activity. It can also improve the efficiency of forest fire detection using remote AI-powered vision analysis.
Key enablers for deploying AI-enabled forest fire management systems
An AI-forest fire management system is predicated on five key enablers. At the core of the system is reliable, localised data on the factors that cause forest fires. In addition, well-functioning systems require sufficient institutional capacity, stakeholder coordination, sustainable financing and community inclusion (Figure 1).
Figure 1: A framework for deploying an AI-enabled forest fire management system
Source: GSMA Mobile for Development
Key findings from our country readiness assessment
A capacity assessment we conducted for the adoption of AI-enabled forest fire management in Pakistan identified numerous challenges as well as possible solutions to address these challenges.
1. Lack of historical data and imagery
Firstly, there is limited availability of historical data and imagery necessary for training AI models. To address the data gap, a crucial initial step is digitising data collection mechanisms within forestry departments and transitioning to Geographic Information System (GIS) based forest monitoring. Fostering mutually beneficial public-private data-sharing collaborations with agritechs and climate tech startups will also help mitigate data limitations. For example, GSMA Innovation Fund grantee, Buraq Integrated Solutions is providing data generated by its weather stations and IoT devices for flood risk and landslide prediction in Pakistan.
There are other companies that could provide valuable data and overcome some of the data gap via public-private partnerships. These include agritech startups like Bakhabar Kissan, another GSMA Innovation Fund grantee, and weather companies such as WeatherWalay, Pakistan’s first private sector player in this space. Proactive engagement from public institutions engaged in forest fire management with such private sector companies could facilitate more technology-enabled DRR initiatives in the country.
2. Limited financial resources
Besides the challenges of data availability and data sharing, Pakistan faces the systemic challenge of limited financial resources for climate risk mitigation and adaptation. Identifying a source of long-term climate-financing for DRR is essential before scaling emerging technologies for forest fire management, if these initiatives are to be sustainable.
3. Constrained human resources
Pakistan’s disaster management authorities and forestry departments also face human resource constraints, both in terms of the number of resources as well as technically skilled resources. Pakistan’s universities produce well-trained data engineers and analysts, making partnerships with academic researchers for algorithm development and testing an excellent starting point for system development. Integrating AI for DRR into higher education curriculums in engineering and ICT, and GIS training within forestry training institutes, will help build long-term capacity for technology-enabled disaster risk management.
An equally important consideration is the need to incorporate communities living in disaster-prone areas such as those susceptible to forest fires, in the design of AI-enabled DRRs. Pakistan’s forest-dwelling communities utilise and protect forest resources, and their involvement in data generation and maintenance of AI-enabled technology deployments in the field could be a key enabler.
Recommendations
Given that Pakistan has not yet embarked on AI-driven DRR, a pragmatic option could be to take a phased approach to developing an AI-enabled forest fire management system. Leveraging available open-access data, such as remotely sensed data from NASA and the European Space Agency, and existing data from domestic sources, could serve as a starting point for algorithm development. Subsequently, additional technologies such as IoT devices and infrared and thermal cameras could be deployed for more reliable, granular and localised data collection to improve the accuracy of the model. AI-enabled forest-fire management systems in Nepal (SERVIR-HKH) and Ethiopia (RISICO) both leverage open-access data and could be replicated in Pakistan.
Adopting emerging technologies for disaster management can be daunting due to the perceived complexity of implementing AI/ML solutions and the need to establish new operating procedures. GSMA’s recently launched report on Combatting forest fires with AI in Pakistan provides a clear roadmap to action, with a view to supporting countries like Pakistan enhance their ability to prevent and mitigate the impact of natural hazards.
Connect with us
The GSMA Mobile for Development’s Central Insights Unit is scaling our work in the area of AI for impact in low- and middle-income countries. In July 2024, we will be publishing an extensive study on AI enablers and use cases in Africa, which includes an analysis of use cases for climate action.
To learn more about our work, you can connect with us at [email protected].
The Central Insights Unit is currently funded by the UK Foreign, Commonwealth & Development Office and supported by the GSMA and its members.
