High Resolution Land Cover Mapping across the Conterminous USA using Deep Learning

Authors: Amin Tayyebi*, ESRI, Daniel Wilson, ESRI, Omar Maher, ESRI, Shairoz Sohail, ESRI
Topics: Global Change, Land Use, United States
Keywords: Land Cover, NAIP Image, Deep Learning, High Resolution, Ecosystem Services
Session Type: Paper
Day: 4/5/2019
Start / End Time: 9:55 AM / 11:35 AM
Room: Washington 6, Marriott, Exhibition Level
Presentation File: No File Uploaded


Global land cover maps have been extensively used for a variety of applications including ecosystem services, climate change, hydrological processes and policy making at local and regional scales. Although low spatial (e.g., 30m) and temporal (e.g., every 5 years) resolution land cover maps have been developed by various agencies (e.g., USGS, USDA, NASA) for the entire Europe and USA, creating high resolution spatial and temporal land cover maps (e.g., 1m) at regional scale is lacking. The land change science community has been pursuing this goal since the early 2000s without success. There are three main issues that need to be addressed to achieve this goal: 1) develop a model to predict multiple land cover classes at regional scale: we used a U-Net based deep learning algorithm with further modifications to develop multi-class land cover model, 2) handle big data: data processing includes getting data from web services, labelling data, training the model, inferencing on new images, mosaicking predicted images, etc. and 3) deploy model at scale for end-users: we are deploying our model in the ArcGIS platform so that users can process NAIP imagery for specific location and get back a classified land cover map. More detailed information about data, model and how users can apply it to their study area will be provided at the conference.

Abstract Information

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

To access contact information login