Project Description

Introduction

Wildfires pose a significant threat in Brazil, causing devastating losses to communities and the economy. A recent incident highlights the gravity of the situation: Gabriel, a farmer in Ribeirão Preto, São Paulo, lost 10 acres of his coffee plantation to wildfires on August 24th. This personal tragedy underscores a broader issue affecting millions of farmers across the country.

upload_image.20:58:02.960694

The Problem

The weather data in the screenshots below show that the fire on Gabriel's farm could have been predicted.

In the weeks leading up to the fire, the region experienced extreme heat, a lack of precipitation, and high wind gusts reaching up to 48 km/h from the northwest. These conditions created a "perfect storm" for wildfires. On August 23rd, four fire outbreaks were recorded near Gabriel's farm, as evidenced by data from NASA's Fire Information for Resource Management System (FIRMS). The following day, the fires escalated, consuming most of his farm despite efforts to contain them.

upload_image.22:04:36.777971

upload_image.22:05:12.725373 upload_image.22:05:50.388569

The economic impact on Gabriel is profound. The loss of 10 acres equates to 500 bags of coffee beans, valued at approximately USD 150,000. Moreover, the damage to his coffee trees means he may not harvest for the next four years, potentially increasing his losses to half a million dollars. Multiply this by the millions of farmers like Gabriel, and the economic ramifications of wildfires become a billion-dollar problem in Brazil.

upload_image.22:07:49.739330

The Gap

Despite the abundance of data on weather patterns and wildfire occurrences, this information often fails to reach those who need it most — farmers like Gabriel. The disconnect arises from a divide between data scientists and agricultural workers. Farmers typically lack access to, or familiarity with, the advanced data tools and platforms that could inform them about impending risks.

upload_image.22:08:46.836031

Our Solution

To bridge this gap, we have developed a voice-powered AI interface that leverages Large Language Models (LLMs) to provide real-time, natural language access to geospatial data. Our platform integrates data from multiple sources, including NASA's FIRMS wildfire database, and presents it through a user-friendly ChatGPT-like interface.

upload_image.22:09:29.851476

How It Works

Farmers can interact with the AI agent using natural language queries. For example, they can ask, "How many wildfires were there in Brazil in the last week?" The AI processes this request by accessing the latest GIS data and provides an accurate, concise answer. The interface supports queries about past events, current conditions, and can potentially offer predictive insights based on weather patterns and historical data.

upload_image.22:10:17.517969

Product demo

Experience the power of our AI Agent in this project demo, showcasing how geospatial data becomes more accessible through a voice-powered, ChatGPT-like interface. This video demonstrates how users can effortlessly interact with complex geospatial information using natural language voice commands.

Product features:

Show how many wildfires there were in the last seven days querying the NASA FIRMS Database per country (currently only Brazil supported)
Show current weather conditions for any location worldwide
Ask the AI Agent to move the map programmtically to a new desired location
Ask about how many days without rain there have been
Ask about context about a country, region, wildfires and other doubts!

Try our project

To try our project, do the following steps:

Access the project link: https://nasa-wildfires-chat-dashboard.netlify.app/
Enter the following API key into the page: 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
Push the connect button, select vad option on the left-bottom side and talk to the AI Agent to get information on wildfires !

Implementation Details

Data Integration: Our system aggregates geospatial data from NASA FIRMS and meteorological sources, ensuring up-to-date information on wildfires and weather conditions.
AI and Natural Language Processing: Utilizing advanced LLMs, the AI agent understands and processes natural language queries, making the interaction intuitive for users without technical expertise.
Voice-Powered Interface: The chat-based interface is accessible via web and mobile platforms, designed with simplicity to accommodate users with varying levels of technological proficiency.
Real-Time Processing: The system provides immediate responses by processing data in real-time, crucial for timely decision-making in emergency situations.
Structured-Output: By querying the LLM and requesting structured output, we are able to programmatically move the map to a new latitude and longitude.
AI-Generated SQL Queries: Based on the user's voice commands, the AI agent has generates SQL queries that can extract relevant information from the database.

Impact

By democratizing access to critical environmental data, our project empowers farmers to make informed decisions to protect their livelihoods. The platform serves as a vital tool for early warning, enabling users to take preventive measures against wildfires. Beyond individual benefits, widespread adoption could significantly reduce the economic losses associated with wildfires in Brazil.

Case Study: Gabriel's Farm

Had Gabriel had access to such a tool, he might have been alerted to the high-risk conditions—the extreme heat, lack of rain, and nearby fire outbreaks. With this information, he could have implemented preventative measures or coordinated with local authorities to mitigate the risk.

Future Work

Predictive Analytics: Incorporate machine learning models to predict wildfire risks based on weather forecasts and historical data.
Expanded Data Sources: Integrate additional datasets, such as vegetation indices and soil moisture levels, for more comprehensive insights.
Collaboration with Authorities: Partner with government agencies and NGOs to disseminate the tool and provide training for effective use.

Conclusion

Our project aims to transform the way farmers interact with critical environmental data. By bridging the gap between complex GIS datasets and accessible communication, we offer a practical solution to a pressing problem. Empowering farmers with timely, relevant information can lead to proactive measures, potentially saving livelihoods and reducing the substantial economic impact of wildfires in Brazil.

2024 NASA Space Apps Challenge

Team I.O.