GIS stands for Geographic Information System; one of its major functionality is to conduct spatial analysis, manually, in the current stage. Large Language Models (LLMs), such as ChatGPT, demonstrate a strong understanding of human natural language and have been explored and applied in various fields, including reasoning, creative writing, code generation, translation, and information retrieval.
By adopting LLM as the reasoning core, we introduce Autonomous GIS, an AI-powered geographic information system (GIS) that leverages the LLM’s general abilities in natural language understanding, reasoning, and coding for addressing spatial problems with automatic spatial data collection, analysis and visualization. We envision that autonomous GIS will need to achieve five autonomous goals including self-generating, self-organizing, self-verifying, self-executing, and self-growing. We developed a prototype system called LLM-Geo using GPT-4 API in a Python environment, demonstrating what an autonomous GIS looks like and how it delivers expected results without human intervention using three case studies.
For the case studies, LLM-Geo successfully returned accurate results, including aggregated numbers, graphs, and maps, significantly reducing manual operation time. Although still lacking several important modules such as logging and code testing, LLM-Geo demonstrates a potential path towards next-generation AI-powered GIS. We advocate for the GIScience community to dedicate more effort to the research and development of autonomous GIS, making spatial analysis easier, faster, and more accessible to a broader audience.
Overall workflow of LLM-Geo
Check out the published paper here: Autonomous GIS: the next-generation AI-powered GIS. Recommended citation format: Li Z., Ning H., 2023. Autonomous GIS: the next-generation AI-powered GIS. Interntional Journal of Digital Earth. https://doi.org/10.1080/17538947.2023.2278895
Note: We are still developing LLM-Geo, and the ideas presented in the paper may change due to the rapid development of AI. For instance, the token limitation appears to have been overcome by Claude (released on 2023-05-11). We hope LLM-Geo can inspire GIScience professionals to further investigate on autonomous GIS.
Clone or download the repository, rename your_config.ini as config.ini. Then, put your OpenAI API key in the config.ini file. Please use GPT-4, the lower versions such as 3.5 do no have enough reasoning ability to generate correct solution graph and operation code.
If you have difficulties installing GeoPandas in Windows, refer to this post.
- Download all files, put your question to the
TASKvariable in LLM-Geo4.ipynb. - Set the
task_namein the notebook. Space is not allowed. LLM-Geo will create the fold using thetask_nameto save results. - Run all cells.
- LLM-Geo will use the backed LLM (GPT-4 now) to review and debug the generated program. GPT-4's debugging ability is still weak. The default maximum attempt count is set to 10; modify this value is needed.
- Note that the solution based on graph, code review, and debug will cost a lot of tokens. We provide a Jupyter notebook (Direct_request_LLM.ipynb) to directly request solutions from LLM. This is a much more quick way to get solutions for simple tasks/questions, but its robustness may be slightly lower.
These case studies are designed to show the concepts of autonomous GIS. Please use GPT-4; the lower version of GPT will fail to generate the correct code and results. Note every time GPT-4 generates different outputs, your results may look different. Per our test, the generated program may not succeed every time, but there is about an 80% chance to run successfully. If input the generated prompts to the ChatGPT-4 chat box rather than API, the success rate will be much higher. We will improve the overall workflow of LLM-Geo, currently we do not push the entire historical conversation (i.e., sufficient information) to the GPT-4 API.
Video demonstrations for the case studies
Case 1: https://youtu.be/ot9oA_6Llys
Case 2: https://youtu.be/ut4XkMcqgvQ
Case 3: https://youtu.be/4q0a9xKk8Ug
This spatial problem is to find out the population living with hazardous wastes and map their distribution. The study area is North Carolina, United States (US). We input the task (question) to LLM-Geo as:
Task:
1) Find out the total population that lives within a tract that contains hazardous waste facilities. The study area is North Carolina, US.
2) Generate a map to show the spatial distribution of the population at the tract level and highlight the borders of tracts that have hazardous waste facilities.
Data locations:
1. NC hazardous waste facility ESRI shape file location: https://github.com/gladcolor/LLM- Geo/raw/master/overlay_analysis/Hazardous_Waste_Sites.zip
2. NC tract boundary shapefile location: https://github.com/gladcolor/LLM-Geo/raw/master/overlay_analysis/tract_shp_37.zip. The tract id column is 'Tract'.
3. NC tract population CSV file location: https://github.com/gladcolor/LLM-Geo/raw/master/overlay_analysis/NC_tract_population.csv. The population is stored in the 'TotalPopulation' column. The tract ID column is 'GEOID'
The results are: (a) Solution graph, (b) assembly program (Python codes), and (c) returned population count and generated map.
NOTE: Please ignore this case since the involved API has been shut down.
This task is to investigate the mobility changes during COVID-19 pandemic in France 2020. First, we asked LLM-Geo to retrieve mobility data from the ODT Explorer using REST API, and then compute and visualize the monthly change rate compared to January 2020. We input the task (question) to LLM-Geo as follows:
Task:
1) Show the monthly change rates of population mobility for each administrative regions in a France map. Each month is a sub-map in a map matrix. The base of the change rate is January 2020.
2) Draw a line chart to show the monthly change rate trends of all administrative regions.
Data locations:
1. ESRI shapefile for France administrative regions: https://github.com/gladcolor/LLM-Geo/raw/master/REST_API/France.zip. The 'GID_1' column is the administrative region code, 'NAME_1' column is the administrative region name.
2. REST API URL with parameters for mobility data access: http://gis.cas.sc.edu/GeoAnalytics/REST?operation=get_daily_movement_for_all_places&source=twitter&scale=world_first_level_admin&begin=01/01/2020&end=12/31/2020. The response is in CSV format. There are three columns in the response: place, date (format:2020-01-07), and intra_movement. 'place' column is the administrative region code, France administrative regions start with 'FRA'.
The results are: (a) Solution graph, (b) map matrix showing the spatial distribution of mobility change rate, (c) line chart showing the trend of the mobility change rate, (d) assembly program.
Note: The ODT explorer API needs to be woken up before being used. Simple open this URL: http://gis.cas.sc.edu/GeoAnalytics/od.html in your browser, then fresh the webpage until you see the flows counts like bellow:
The spatial problem for this case is to investigate the spatial distribution of the COVID-19 death rate (ratio of COVID-19 deaths to cases) and the association between the death rate and the proportion of senior residents (age >=65) at the US county level. The death rate is derived from the accumulated COVID-19 data as of December 31, 2020, available from New York Times (2023), based on state and local health agency reports. The population data is extracted from the 2020 ACS five-year estimates (US Census Bureau 2022). The task asks for a map to show the county level death rate distribution and a scatter plot to show the correlation and trend line of the death rate with the senior resident rate. We input the task (question) to LLM-Geo as:
Task:
1) Draw a map to show the death rate (death/case) of COVID-19 among the contiguous US counties. Use the accumulated COVID-19 data of 2020.12.31 to compute the death rate. Use scheme ='quantiles' when plotting the map. Set map projection to 'Conus Albers'. Set map size to 15*10 inches.
2) Draw a scatter plot to show the correlation and trend line of the death rate with the senior resident rate, including the r-square and p-value. Set data point transparency to 50%, regression line as red. Set figure size to 15*10 inches.
Data locations:
1) COVID-19 data case in 2020 (county-level): https://github.com/nytimes/covid-19-data/raw/master/us-counties-2020.csv. This data is for daily accumulated COVID cases and deaths for each county in the US. There are 5 columns: date (format: 2021-02-01), county, state, fips, cases, deaths.
2) Contiguous US county boundary (ESRI shapefile): https://github.com/gladcolor/spatial_data/raw/master/contiguous_counties.zip. The county FIPS column is 'GEOID'.
3) Census data (ACS2020): https://raw.githubusercontent.com/gladcolor/spatial_data/master/Demography/ACS2020_5year_county.csv. The needed columns are: 'FIPS', 'Total Population', 'Total Population: 65 to 74 Years', 'Total Population: 75 to 84 Years', 'Total Population: 85 Years and Over'.
The results are: (a) Solution graph, (b) county level death rate map of the contiguous US, (c) scatter plot showing the association between COVID-19 death rate and the senior resident rate at the county level, (d) assembly program.
Understand the architecture of LLM-Geo might help you customize it or develop you own autonomous GIS agents; here is a brief introduction of the architecture.
- Test with more case studies.
- Improve the prompt generation.
- Implement an autonomous data understanding module.
- Implement an autonomous data collection module. (Working on another data download agent)
- Implement an autonomous data visualization module. (Working on another cartograph agent)
- Develop a web-based front-end user interface.
- Test the integration of LLM-Geo with QGIS.
- Integrate and evaluate the open source LLM Meta Llama 3 (8B and 70B) as an alternative reasoning core in LLM-Geo.
- You may need the geopandas package to load vector files. Please install it in advance.