Artificial Intelligence

The replicability and reproducibility challenges accompanying the calculation of environmental land surface parameters

Friday, October 12, 2018, 11:00am - 12:00pm PDTiCal
6th floor large conference room
This event is open to the public.
AI Seminar
John P. Wilson, USC

The land surface plays a fundamental role in modulating several of the Earth’s dynamic systems including a large number of atmospheric, geologic, geomorphic, hydrologic, and ecological processes. Consequently, there is a growing interest in quantitatively characterizing the land surface and segmenting the topography into fundamental spatial units, as the topography inherently represents the results of the interplay between various systems, and records an imprint of landscape dynamics.

Applications that exploit this knowledge usually rely on Digital Elevation Models (DEMs) to represent the surface and a steadily increasing and sophisticated range of techniques for topographic analysis, modeling, and visualization. Many of these innovations have accompanied the rapid proliferation of geographic information technologies, which has provided new data, algorithms, analysis, and modeling techniques for characterizing the Earth’s surface. These techniques and the accompanying digital data have fueled the rise of geomorphometry, which in its broadest sense, refers to the science of quantitative land surface characterization.

Modern geomorphometry focuses on the extraction of land surface parameters and segmentation of the landscape into spatial entities or features (i.e. land surface objects) from digital topography. Many questions still remain, and users must be aware of the advantages and disadvantages associated with various representations and data structures, metrics and indices, spatial modeling approaches, and their utility for scientific investigations. Most of these questions can be attributed to the steady growth in the number of parameters and algorithms for processing DEMs and extracting both the descriptive measures (parameters) and surface features (objects). The values of these parameters and the characteristics of the objects will vary depending on a variety of factors, including the landscape at hand, parameterization scheme, measurement scale of the data, mathematical model by which they are calculated, size of the search window, and the grid resolution.

This presentation will first review the calculation of the slope direction and flow accumulation land surface parameters, which are important for routing flow across the land surface and second, to highlight the replicability and reproducibility challenges that accompany the calculation of these land surface parameters and how modern computational approaches might help us to resolve such issues.

Dr. John P. Wilson is Professor of Sociology and Spatial Sciences in the USC Dana and David Dornsife College of Letters, Arts and Sciences where he directs the Spatial Sciences Institute as well as the Wilson Map Lab. He also serves as GIS Lead for the Spatial and Exposure Analytics Core in the Southern California Environmental Health Sciences Center, and holds courtesy appointments as Professor in the School of Architecture, in the Keck School of Medicine of USC’s Department of Preventive Medicine, and in the Viterbi School of Engineering’s Departments of Computer Science and Civil & Environmental Engineering. His research focuses on the modeling of human and environmental systems and makes extensive use of GIS, spatial analysis, and computer models. He has published numerous books and articles on these topics, including Environmental Applications of Digital Terrain Modeling (Wiley-Blackwell, 2018) and two edited volumes, Terrain Analysis: Principles and Applications (Wiley, 2000) and The Handbook of Geographic Information Science (Blackwell, 2008). Much of this work is collaborative and cross-disciplinary with the goal of improving our understanding of the factors linking people, their environments, and their health. The work of his lab group can be seen at

« Return to Events