GeoGPT+: Using OpenAI’s custom GPTs for geospatial analysis

Introduction

In our fast-paced world, quick and accurate decision-making is often hindered by the complexities and time-consuming nature of geospatial analysis. Despite the advancements in Geographic Information Systems (GIS), the access to and processing of geospatial data remain complex tasks, especially for non-experts.

Current Challenges

As observed by Manvi et al. (2023), “In a fast-paced world where decisions unfold at lightning speed, accessing accurate and timely geospatial data remains a challenge.” Moreover, Zhang et al. (2023) rightly points out that “Geospatial professionals can easily handle geospatial tasks using sequential tools, spatial algorithms and GIS operations, it’s difficult for non-experts.”

Adding to these challenges, Mai et al. (2023) highlights the complexity in developing GeoAI language models (LLMs), particularly due to the multimodal nature of geospatial tasks.

The main challenges we have experienced:

• High costs and complexity of existing GIS tools.
• Non-experts struggling with the use of these tools.
• Time-intensive processes in geospatial analysis.

Our Solution: GeoGPT

In response to these challenges, our team has developed GeoGPT — a solution leveraging generative AI to automate data collection and analysis. This web-hosted tool enables users to generate geospatial data and conduct analysis through text prompts.

ChatGPT, developed by OpenAI, made a significant impact globally with its progressive versions released between 2019 and 2023. The introduction of “GPTs” (customizable ChatGPT versions) and the GPT store in 2024 marked another milestone. GeoGPT, a concept that emerged while ChatGPT still operated on a plugin system, was later developed under the new GPT framework.

Our idea for developing GeoGPT, originated during the Near-East conflict in 2023, addressing the need for rapid, reliable batch geocoding to identify impacted communities and extract data to support humanitarian responses. Our team recognizes the challenges in accessing and processing relevant datasets, especially for those unfamiliar with geospatial concepts.

By the time GPT-3.5 was released, there were already discussions about a Geo-LLM, and several research papers had explored the idea. The enhancements to ChatGPT, including its ability to process JSON data, interpret images, and utilize APIs, made the development of a public Geo-LLM feasible.

Our initial experiments using OpenAI’s API on a local machine yielded promising results, encouraging us to create a specialized GPT integrating diverse datasets, such as administrative boundaries (levels 0, 1, and 2), high-resolution satellite imagery (10m), building footprints, population data, NDVI from satellite imagery, and geocoding capabilities including the ability to extract and geocode location data from images.

GeoGPT is designed for both speed and ease of use, aiming to attract a wide user base and reduce data extraction and processing times.

Case Study: Humanitarian application in the Near-East Conflict

In a case study during the Near-East conflict in October, we utilized GeoGPT to geocode communities evacuated in Israel. By simply uploading a screenshot from the IDF’s telegram channel, we received a list of evacuated communities.

This tool also enables users to mark these communities on a map and export their coordinates in a table or CSV file, streamlining the process by eliminating manual coordinate entry and facilitating integration with existing GIS software.

Future Prospects

This is only the beginning stages of a very exciting project however, and we hope to add functionalities and datasets as time goes on.

Learn More and Get Involved

For more details about GeoGPT and its features, visit our website at ageospatial.com or watch our demo video. You can also try GeoGPT+ (requires ChatGPT 4). We welcome your feedback and support, and you can reach out to us through our contact form or email (contact@ageospatial.com).

Support Our Work

Your support is invaluable to us. If you wish to contribute, you can support our work by buying us a coffee.