Predicting future farmland pattern using GeoAI -- A case study of Changzhou, China

Authors: Qiuhao Huang*, Nanjing University, Linlin Sheng, Nanjing Normal University, China
Topics: Land Use
Keywords: farmland, modeling, GIS, Machine learning, Changzhou
Session Type: Paper
Day: 4/6/2019
Start / End Time: 5:00 PM / 6:40 PM
Room: 8217, Park Tower Suites, Marriott, Lobby Level
Presentation File: No File Uploaded

Farmland is precious resources in maintaining food security and ecological security. In the past decades, China experienced farmland loss drastically, especially around urban-rural fringes in Eastern China. The key to adequately protect the farmland is to predict future farmland pattern and take countermeasures in advance. This paper attempts to predict future farmland pattern combining GIS technology and Machine Learning (ML) algorithms. Changzhou city, a fast-growing city located in Yangtze River Delta, is selected as the study area. Firstly, we divided the study area into 100m*100m fishnet. Then, nine variables that correlate with the farmland distribution are collected. Using the GIS spatial analysis, we assigned the value of the nine variables to the spatially corresponding fishnets. Later, different ML algorithms were tried using the fishnet data to select the better models. Through model parameters tuning and model validation, we get the satisfactory models for farmland pattern. Finally, we identify the importance of the models’ factors and predict the farmland pattern in 2025. This paper not only supports the local urban planner and land use managers for farmland protection decisions, but also provides an innovative method for land use modeling.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login